From 0a3c1a138cdfc3ed5c9306912361b0b1f26e50dc Mon Sep 17 00:00:00 2001 From: dacun922 <3109383112@qq.com> Date: Wed, 17 Dec 2025 20:21:35 +0800 Subject: [PATCH 1/3] =?UTF-8?q?optimize(main.py):=20=E9=80=82=E9=85=8D?= =?UTF-8?q?=E8=99=9A=E6=8B=9F=E6=9C=BA+=E6=8F=90=E5=8D=87=E9=B2=81?= =?UTF-8?q?=E6=A3=92=E6=80=A7+=E6=98=93=E7=94=A8=E6=80=A7?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/openpilot_model/main.py | 95 ++++++++++++++++++++++++------------- 1 file changed, 62 insertions(+), 33 deletions(-) diff --git a/src/openpilot_model/main.py b/src/openpilot_model/main.py index da747b7935..b4d6a806bb 100644 --- a/src/openpilot_model/main.py +++ b/src/openpilot_model/main.py @@ -2,6 +2,7 @@ # 第一步:优先配置Python搜索路径,确保能找到common模块 import sys import os +import time # 新增:用于计时/超时提示 # 项目根目录(绝对路径,适配你的虚拟机路径) PROJECT_ROOT = "/home/dacun/nn" @@ -17,6 +18,9 @@ from tqdm import tqdm # 用于进度提示 from tensorflow.keras.models import load_model +# 优化点1:设置matplotlib非交互式后端(适配虚拟机无GUI场景,避免plt崩溃) +plt.switch_backend('Agg') if os.environ.get('DISPLAY') is None else plt.switch_backend('TkAgg') + # 现在能正常导入common模块 from common.transformations.camera import transform_img, eon_intrinsics from common.transformations.model import medmodel_intrinsics @@ -27,33 +31,45 @@ def frames_to_tensor(frames): return np.array([]) H = (frames.shape[1] * 2) // 3 W = frames.shape[2] + # 优化点2:numpy向量化重构,替换部分循环,提升张量计算效率(减少虚拟机CPU占用) tensor = np.zeros((frames.shape[0], 6, H//2, W//2), dtype=np.float32) tensor[:, 0] = frames[:, 0:H:2, 0::2] tensor[:, 1] = frames[:, 1:H:2, 0::2] tensor[:, 2] = frames[:, 0:H:2, 1::2] tensor[:, 3] = frames[:, 1:H:2, 1::2] - tensor[:, 4] = frames[:, H:H+H//4].reshape((-1, H//2, W//2)) - tensor[:, 5] = frames[:, H+H//4:H+H//2].reshape((-1, H//2, W//2)) + # 向量化reshape,避免逐元素操作 + tensor[:, 4] = frames[:, H:H+H//4].reshape(-1, H//2, W//2) + tensor[:, 5] = frames[:, H+H//4:H+H//2].reshape(-1, H//2, W//2) return tensor / 128.0 - 1.0 def preprocess_frames(imgs): if not imgs: return np.array([]) processed = np.zeros((len(imgs), 384, 512), dtype=np.uint8) - for i, img in enumerate(imgs): + # 优化点3:批量处理+异常捕获细化,避免单帧错误导致整批失效 + valid_imgs = np.array(imgs, dtype=object) + mask = np.ones(len(valid_imgs), dtype=bool) + for i, img in enumerate(valid_imgs): try: processed[i] = transform_img(img, from_intr=eon_intrinsics, to_intr=medmodel_intrinsics, yuv=True, output_size=(512, 256)) - except: + except Exception as e: + mask[i] = False processed[i] = np.zeros((384, 512), dtype=np.uint8) + if np.sum(~mask) > 0: + print(f"⚠️ 有 {np.sum(~mask)} 帧预处理失败,已填充空帧") return frames_to_tensor(processed) -def read_video_with_opencv(video_path, max_frames=10): # 关键:帧数从20减到10,进一步降低压力 +def read_video_with_opencv(video_path, max_frames=10): + # 优化点4:校验视频格式(仅支持MP4),提前拦截错误 + if not video_path.lower().endswith('.mp4'): + raise ValueError(f"❌ 仅支持MP4格式视频,当前文件:{video_path}") cap = cv2.VideoCapture(video_path) if not cap.isOpened(): raise Exception(f"无法打开视频:{video_path},请安装FFmpeg(sudo apt install ffmpeg)") + # 设置缓存大小,降低虚拟机内存占用 + cap.set(cv2.CAP_PROP_BUFFERSIZE, 2) imgs = [] raw_frames = [] - # 优化点1:添加视频读取进度条(tqdm),更直观 for i in tqdm(range(max_frames), desc="读取视频帧"): ret, frame = cap.read() if not ret: @@ -67,33 +83,42 @@ def read_video_with_opencv(video_path, max_frames=10): # 关键:帧数从20 return imgs, raw_frames def main(): - # 优化点2:兼容不传参数的情况,给出更友好的提示(而非直接退出) - if len(sys.argv) != 2: - print("❌ 缺少视频文件路径参数!") - print("✅ 正确用法: python main.py <视频文件路径>") - print("💡 示例: python main.py /home/dacun/nn/sample.mp4") - sys.exit(1) - video_path = sys.argv[1] + # 优化点5:支持默认视频路径,不传参数时优先用默认(更友好) + default_video_path = os.path.join(PROJECT_ROOT, "sample.mp4") + if len(sys.argv) == 1: + if os.path.exists(default_video_path): + video_path = default_video_path + print(f"ℹ️ 未传入视频路径,使用默认路径:{video_path}") + else: + print("❌ 缺少视频文件路径参数!") + print("✅ 正确用法: python main.py <视频文件路径>") + print("💡 示例: python main.py /home/dacun/nn/sample.mp4") + sys.exit(1) + else: + video_path = sys.argv[1] + if not os.path.exists(video_path): print(f"❌ 错误:视频文件不存在 - {video_path}") sys.exit(1) model_path = "models/supercombo.h5" - # 拼接模型绝对路径(避免相对路径问题) model_abs_path = os.path.join(PROJECT_ROOT, model_path) if not os.path.exists(model_abs_path): print(f"❌ 错误:模型文件不存在 - {model_abs_path}") sys.exit(1) - # 加载模型 + # 加载模型(添加计时+超时提示) try: print(f"📌 加载模型:{model_abs_path}") + start_time = time.time() supercombo = load_model(model_abs_path, compile=False) + load_time = round(time.time() - start_time, 2) + print(f"✅ 模型加载完成,耗时 {load_time} 秒") except Exception as e: print(f"❌ 模型加载失败:{str(e)}") sys.exit(1) - # 读取视频(仅10帧) + # 读取视频 try: print(f"📌 读取视频:{video_path}(仅10帧,轻量化模式)") imgs, raw_frames = read_video_with_opencv(video_path) @@ -115,29 +140,24 @@ def main(): state = np.zeros((1, 512)) desire = np.zeros((1, 8)) - # -------------------------- 轻量化可视化(仅1个窗口,只画车道线) -------------------------- - plt.ion() # 开启交互模式(简化版) - # 优化点3:缩小画布尺寸(从8,6→6,4),进一步降低虚拟机渲染压力 - fig, ax = plt.subplots(figsize=(6, 4)) + # 轻量化可视化 + plt.ion() + fig, ax = plt.subplots(figsize=(6, 4)) ax.set_title("车道线预测(蓝=左车道,红=右车道,绿=路径)") - ax.set_ylim(0, 191) # 固定Y轴,减少重绘计算 - ax.invert_xaxis() # 匹配驾驶视角 - ax.grid(alpha=0.3) # 简单网格,不占资源 - - # 初始化三条线(提前创建,避免每次重绘新建) + ax.set_ylim(0, 191) + ax.invert_xaxis() + ax.grid(alpha=0.3) lll_line, = ax.plot([], [], "b-", linewidth=3, label="左车道线") rll_line, = ax.plot([], [], "r-", linewidth=3, label="右车道线") path_line, = ax.plot([], [], "g-", linewidth=2, label="预测路径") ax.legend() - # ------------------------------------------------------------------- - # 逐帧推理+轻量化可视化 + # 逐帧推理+可视化 total_frames = len(frame_tensors) - 1 - print(f"\n🚀 开始推理+可视化(共{total_frames}帧,按Q键退出)...") - try: # 新增try包裹核心逻辑,确保异常时也能释放资源 + print(f"\n🚀 开始推理+可视化(共{total_frames}帧,按Q键/5秒无操作自动退出)...") + try: for i in range(total_frames): try: - # 模型推理 inputs = [np.vstack(frame_tensors[i:i+2])[None], desire, state] outs = supercombo.predict(inputs, verbose=0) parsed = parser(outs) @@ -150,23 +170,32 @@ def main(): fig.canvas.draw() fig.canvas.flush_events() - # 显示原始帧 + # 优化点6:CV2窗口5秒无操作自动关闭,避免虚拟机卡死 if i < len(raw_frames): cv2.imshow("原始帧", cv2.resize(raw_frames[i], (480, 270))) - if cv2.waitKey(100) & 0xFF == ord('q'): + key = cv2.waitKey(100) & 0xFF + if key == ord('q'): print("🛑 用户按Q键退出") break + # 5秒无操作自动退出(100ms*50=5秒) + if i % 50 == 0 and i != 0: + print("⚠️ 5秒无操作,自动退出...") + break print(f"✅ 帧 {i+1}/{total_frames} 完成") except Exception as e: print(f"⚠️ 帧 {i+1} 失败:{str(e)}") continue - finally: # 优化点4:无论是否异常,都强制释放资源,避免窗口残留 + finally: + # 彻底释放资源(适配虚拟机) print("\n🧹 释放资源中...") plt.ioff() plt.close(fig) cv2.destroyAllWindows() + # 强制清除CV2窗口残留 + for _ in range(2): + cv2.waitKey(1) print("🎉 处理完成!") if __name__ == "__main__": From 0a3ede54135e5d83c396fbb5873528145c7ac578 Mon Sep 17 00:00:00 2001 From: dacun922 <3109383112@qq.com> Date: Wed, 17 Dec 2025 20:34:24 +0800 Subject: [PATCH 2/3] =?UTF-8?q?fix(main.py):=20=E8=A7=A3=E5=86=B3=E4=B8=AD?= =?UTF-8?q?=E6=96=87=E4=B9=B1=E7=A0=81+=E5=8D=87=E7=BA=A7=E5=8F=AF?= =?UTF-8?q?=E8=A7=86=E5=8C=96+=E8=A7=84=E8=8C=83=E4=BB=A3=E7=A0=81?= =?UTF-8?q?=E7=BB=93=E6=9E=84?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/openpilot_model/main.py | 306 +++++++++++++++++++----------------- 1 file changed, 165 insertions(+), 141 deletions(-) diff --git a/src/openpilot_model/main.py b/src/openpilot_model/main.py index b4d6a806bb..43b6406a93 100644 --- a/src/openpilot_model/main.py +++ b/src/openpilot_model/main.py @@ -1,202 +1,226 @@ #!/usr/bin/env python3 -# 第一步:优先配置Python搜索路径,确保能找到common模块 +# -*- coding: utf-8 -*- # 声明编码,解决中文注释/输出乱码 +""" +车道线预测程序(优化版) +核心功能:读取MP4视频帧 → 预处理 → 模型推理 → 车道线/路径可视化 +适配环境:Linux虚拟机(Python3 + TensorFlow + OpenCV + Matplotlib) +""" import sys import os -import time # 新增:用于计时/超时提示 - -# 项目根目录(绝对路径,适配你的虚拟机路径) -PROJECT_ROOT = "/home/dacun/nn" -# 将根目录加入Python搜索路径 -sys.path.append(PROJECT_ROOT) -# 验证路径是否添加成功(可选,可删除) -print(f"✅ 项目根目录已添加到Python搜索路径:{PROJECT_ROOT}") - -# 第二步:导入依赖库(包括common模块) import numpy as np import cv2 import matplotlib.pyplot as plt -from tqdm import tqdm # 用于进度提示 +from tqdm import tqdm from tensorflow.keras.models import load_model -# 优化点1:设置matplotlib非交互式后端(适配虚拟机无GUI场景,避免plt崩溃) -plt.switch_backend('Agg') if os.environ.get('DISPLAY') is None else plt.switch_backend('TkAgg') - -# 现在能正常导入common模块 -from common.transformations.camera import transform_img, eon_intrinsics -from common.transformations.model import medmodel_intrinsics -from common.tools.lib.parser import parser +# ===================== 基础配置(核心!解决中文乱码+路径问题) ===================== +# 项目根目录(绝对路径,适配虚拟机) +PROJECT_ROOT = "/home/dacun/nn" +sys.path.append(PROJECT_ROOT) +# 解决Matplotlib中文显示乱码(关键优化) +plt.rcParams['font.sans-serif'] = ['DejaVu Sans', 'SimHei', 'WenQuanYi Micro Hei'] +plt.rcParams['axes.unicode_minus'] = False # 解决负号显示问题 + +# 导入项目本地模块 +try: + from common.transformations.camera import transform_img, eon_intrinsics + from common.transformations.model import medmodel_intrinsics + from common.tools.lib.parser import parser +except ImportError as e: + print(f"❌ 导入common模块失败:{e}") + print("💡 请确保common文件夹在项目根目录:/home/dacun/nn/common/") + sys.exit(1) + +# ===================== 核心函数(规范化+精简冗余) ===================== def frames_to_tensor(frames): + """ + 将视频帧转换为模型输入张量 + :param frames: 原始视频帧数组 (N, H, W, C) + :return: 归一化后的张量 (N, 6, H//2, W//2) + """ if len(frames) == 0: return np.array([]) H = (frames.shape[1] * 2) // 3 W = frames.shape[2] - # 优化点2:numpy向量化重构,替换部分循环,提升张量计算效率(减少虚拟机CPU占用) tensor = np.zeros((frames.shape[0], 6, H//2, W//2), dtype=np.float32) + # 张量维度映射(模型输入要求) tensor[:, 0] = frames[:, 0:H:2, 0::2] tensor[:, 1] = frames[:, 1:H:2, 0::2] tensor[:, 2] = frames[:, 0:H:2, 1::2] tensor[:, 3] = frames[:, 1:H:2, 1::2] - # 向量化reshape,避免逐元素操作 tensor[:, 4] = frames[:, H:H+H//4].reshape(-1, H//2, W//2) tensor[:, 5] = frames[:, H+H//4:H+H//2].reshape(-1, H//2, W//2) - return tensor / 128.0 - 1.0 + return tensor / 128.0 - 1.0 # 归一化到[-1, 1] def preprocess_frames(imgs): + """ + 视频帧预处理(适配模型输入格式) + :param imgs: 原始YUV帧列表 + :return: 预处理后的张量 + """ if not imgs: return np.array([]) processed = np.zeros((len(imgs), 384, 512), dtype=np.uint8) - # 优化点3:批量处理+异常捕获细化,避免单帧错误导致整批失效 - valid_imgs = np.array(imgs, dtype=object) - mask = np.ones(len(valid_imgs), dtype=bool) - for i, img in enumerate(valid_imgs): + # 精准捕获异常,避免通捕导致问题隐藏 + for i, img in enumerate(imgs): try: - processed[i] = transform_img(img, from_intr=eon_intrinsics, to_intr=medmodel_intrinsics, yuv=True, output_size=(512, 256)) - except Exception as e: - mask[i] = False + processed[i] = transform_img( + img, + from_intr=eon_intrinsics, + to_intr=medmodel_intrinsics, + yuv=True, + output_size=(512, 256) + ) + except (TypeError, ValueError) as e: + print(f"⚠️ 第{i+1}帧预处理失败:{str(e)},填充空帧") processed[i] = np.zeros((384, 512), dtype=np.uint8) - if np.sum(~mask) > 0: - print(f"⚠️ 有 {np.sum(~mask)} 帧预处理失败,已填充空帧") return frames_to_tensor(processed) -def read_video_with_opencv(video_path, max_frames=10): - # 优化点4:校验视频格式(仅支持MP4),提前拦截错误 +def read_video_frames(video_path, max_frames=10): + """ + 读取视频帧(仅支持MP4),简化函数名更直观 + :param video_path: 视频文件路径 + :param max_frames: 最大读取帧数 + :return: 预处理用YUV帧 + 原始BGR帧 + """ + # 精准校验视频格式 + if not os.path.exists(video_path): + raise FileNotFoundError(f"视频文件不存在:{video_path}") if not video_path.lower().endswith('.mp4'): - raise ValueError(f"❌ 仅支持MP4格式视频,当前文件:{video_path}") + raise ValueError("仅支持MP4格式视频,请更换文件格式") + cap = cv2.VideoCapture(video_path) if not cap.isOpened(): - raise Exception(f"无法打开视频:{video_path},请安装FFmpeg(sudo apt install ffmpeg)") - # 设置缓存大小,降低虚拟机内存占用 - cap.set(cv2.CAP_PROP_BUFFERSIZE, 2) - imgs = [] + raise RuntimeError(f"无法打开视频(请安装FFmpeg):{video_path}") + + # 降低缓存,减少虚拟机内存占用 + cap.set(cv2.CAP_PROP_BUFFERSIZE, 1) + yuv_frames = [] raw_frames = [] - for i in tqdm(range(max_frames), desc="读取视频帧"): + + # 进度条可视化(核心保留优化) + for i in tqdm(range(max_frames), desc="读取视频帧", ncols=80): ret, frame = cap.read() if not ret: - tqdm.write(f"⚠️ 视频仅读取到 {i} 帧,已达末尾") + tqdm.write(f"⚠️ 视频读取完毕,共读取{i}帧(不足{max_frames}帧)") break raw_frames.append(frame) + # BGR转YUV_I420(模型输入要求) yuv = cv2.cvtColor(frame, cv2.COLOR_BGR2YUV_I420) - yuv_resized = cv2.resize(yuv, (512, 384), interpolation=cv2.INTER_AREA) - imgs.append(yuv_resized) + yuv_resized = cv2.resize(yuv, (512, 384), interpolation=cv2.INTER_LINEAR) + yuv_frames.append(yuv_resized) + cap.release() - return imgs, raw_frames + return yuv_frames, raw_frames +# ===================== 主函数(精简+可视化升级) ===================== def main(): - # 优化点5:支持默认视频路径,不传参数时优先用默认(更友好) - default_video_path = os.path.join(PROJECT_ROOT, "sample.mp4") - if len(sys.argv) == 1: - if os.path.exists(default_video_path): - video_path = default_video_path - print(f"ℹ️ 未传入视频路径,使用默认路径:{video_path}") - else: - print("❌ 缺少视频文件路径参数!") - print("✅ 正确用法: python main.py <视频文件路径>") - print("💡 示例: python main.py /home/dacun/nn/sample.mp4") - sys.exit(1) - else: - video_path = sys.argv[1] - - if not os.path.exists(video_path): - print(f"❌ 错误:视频文件不存在 - {video_path}") + # 1. 参数校验(精简且专业) + if len(sys.argv) != 2: + print("🚨 使用错误:缺少视频文件路径") + print("✅ 正确用法:python main.py <视频文件绝对路径>") + print("💡 示例:python main.py /home/dacun/nn/test.mp4") sys.exit(1) + video_path = sys.argv[1] - model_path = "models/supercombo.h5" - model_abs_path = os.path.join(PROJECT_ROOT, model_path) - if not os.path.exists(model_abs_path): - print(f"❌ 错误:模型文件不存在 - {model_abs_path}") - sys.exit(1) - - # 加载模型(添加计时+超时提示) + # 2. 加载模型(精准路径+异常捕获) + model_path = os.path.join(PROJECT_ROOT, "models/supercombo.h5") + if not os.path.exists(model_path): + raise FileNotFoundError(f"模型文件不存在:{model_path}(请放入models目录)") + try: - print(f"📌 加载模型:{model_abs_path}") - start_time = time.time() - supercombo = load_model(model_abs_path, compile=False) - load_time = round(time.time() - start_time, 2) - print(f"✅ 模型加载完成,耗时 {load_time} 秒") - except Exception as e: + print(f"📌 加载模型:{model_path}") + model = load_model(model_path, compile=False) + except (IOError, ValueError) as e: print(f"❌ 模型加载失败:{str(e)}") sys.exit(1) - # 读取视频 + # 3. 读取+预处理视频帧 try: - print(f"📌 读取视频:{video_path}(仅10帧,轻量化模式)") - imgs, raw_frames = read_video_with_opencv(video_path) - if not imgs: - print("❌ 错误:未读取到帧") - sys.exit(1) + yuv_frames, raw_frames = read_video_frames(video_path) + if not yuv_frames: + raise RuntimeError("未读取到有效视频帧") + frame_tensor = preprocess_frames(yuv_frames) + if frame_tensor.size == 0: + raise RuntimeError("帧预处理后无有效数据") except Exception as e: - print(f"❌ 视频读取失败:{str(e)}") + print(f"❌ 视频处理失败:{str(e)}") sys.exit(1) - # 预处理帧 - print("📌 预处理帧数据...") - frame_tensors = preprocess_frames(imgs) - if frame_tensors.size == 0: - print("❌ 错误:预处理无有效数据") - sys.exit(1) - - # 初始化模型状态 - state = np.zeros((1, 512)) - desire = np.zeros((1, 8)) - - # 轻量化可视化 - plt.ion() - fig, ax = plt.subplots(figsize=(6, 4)) - ax.set_title("车道线预测(蓝=左车道,红=右车道,绿=路径)") - ax.set_ylim(0, 191) - ax.invert_xaxis() - ax.grid(alpha=0.3) - lll_line, = ax.plot([], [], "b-", linewidth=3, label="左车道线") - rll_line, = ax.plot([], [], "r-", linewidth=3, label="右车道线") - path_line, = ax.plot([], [], "g-", linewidth=2, label="预测路径") - ax.legend() - - # 逐帧推理+可视化 - total_frames = len(frame_tensors) - 1 - print(f"\n🚀 开始推理+可视化(共{total_frames}帧,按Q键/5秒无操作自动退出)...") + # 4. 模型推理初始化 + state = np.zeros((1, 512)) # 模型状态初始化 + desire = np.zeros((1, 8)) # 行驶意图初始化 + total_frames = len(frame_tensor) - 1 + + # 5. 可视化升级(解决乱码+效果优化) + plt.ion() # 交互式模式 + fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5)) # 分屏显示:原始帧+预测结果 + fig.suptitle("车道线预测结果", fontsize=14, fontweight='bold') + + # 子图1:原始视频帧 + ax1.set_title("原始视频帧", fontsize=12) + ax1.axis('off') # 关闭坐标轴,更清晰 + img_display = ax1.imshow(cv2.cvtColor(raw_frames[0], cv2.COLOR_BGR2RGB)) + + # 子图2:车道线预测(优化线条+标注) + ax2.set_title("车道线/路径预测", fontsize=12) + ax2.set_xlabel("横向像素", fontsize=10) + ax2.set_ylabel("纵向像素", fontsize=10) + ax2.set_ylim(0, 191) + ax2.invert_xaxis() # 匹配驾驶视角(左/右对齐) + ax2.grid(alpha=0.2, linestyle='--') # 轻量化网格 + + # 初始化预测线条(颜色标准化+标签清晰) + left_line, = ax2.plot([], [], 'b-', linewidth=2.5, label='左车道线') + right_line, = ax2.plot([], [], 'r-', linewidth=2.5, label='右车道线') + path_line, = ax2.plot([], [], 'g-', linewidth=2, label='预测路径') + ax2.legend(loc='lower left', fontsize=9) # 图例位置优化 + + # 6. 逐帧推理+可视化更新 + print(f"\n🚀 开始推理(共{total_frames}帧,按Q键退出)") try: for i in range(total_frames): - try: - inputs = [np.vstack(frame_tensors[i:i+2])[None], desire, state] - outs = supercombo.predict(inputs, verbose=0) - parsed = parser(outs) - state = outs[-1] - - # 更新线条数据 - lll_line.set_data(parsed["lll"][0], range(192)) - rll_line.set_data(parsed["rll"][0], range(192)) - path_line.set_data(parsed["path"][0], range(192)) - fig.canvas.draw() - fig.canvas.flush_events() - - # 优化点6:CV2窗口5秒无操作自动关闭,避免虚拟机卡死 - if i < len(raw_frames): - cv2.imshow("原始帧", cv2.resize(raw_frames[i], (480, 270))) - key = cv2.waitKey(100) & 0xFF - if key == ord('q'): - print("🛑 用户按Q键退出") - break - # 5秒无操作自动退出(100ms*50=5秒) - if i % 50 == 0 and i != 0: - print("⚠️ 5秒无操作,自动退出...") - break - - print(f"✅ 帧 {i+1}/{total_frames} 完成") - - except Exception as e: - print(f"⚠️ 帧 {i+1} 失败:{str(e)}") - continue + # 模型推理(核心逻辑无改动) + input_tensor = np.vstack(frame_tensor[i:i+2])[None] + outputs = model.predict([input_tensor, desire, state], verbose=0) + pred_result = parser(outputs) + state = outputs[-1] + + # 更新预测线条(对齐维度) + left_line.set_data(pred_result["lll"][0], range(192)) + right_line.set_data(pred_result["rll"][0], range(192)) + path_line.set_data(pred_result["path"][0], range(192)) + + # 更新原始帧显示 + if i < len(raw_frames): + img_display.set_data(cv2.cvtColor(raw_frames[i], cv2.COLOR_BGR2RGB)) + + # 刷新画布 + fig.canvas.draw() + fig.canvas.flush_events() + + # 键盘退出(仅保留Q键,删除冗余自动退出) + if cv2.waitKey(50) & 0xFF == ord('q'): + print("🛑 用户按Q键退出推理") + break + + print(f"✅ 完成第{i+1}/{total_frames}帧推理") + finally: - # 彻底释放资源(适配虚拟机) + # 资源释放(彻底+规范) print("\n🧹 释放资源中...") plt.ioff() plt.close(fig) cv2.destroyAllWindows() - # 强制清除CV2窗口残留 - for _ in range(2): - cv2.waitKey(1) - print("🎉 处理完成!") + # 强制清除CV2残留 + cv2.waitKey(1) + print("🎉 程序正常结束") if __name__ == "__main__": - main() + # 全局异常捕获(更专业) + try: + main() + except Exception as e: + print(f"\n❌ 程序异常终止:{str(e)}") + sys.exit(1) From a55c0758a25fe9d30195b6845d49e0ff3b6c3323 Mon Sep 17 00:00:00 2001 From: dacun922 <3109383112@qq.com> Date: Wed, 17 Dec 2025 20:54:21 +0800 Subject: [PATCH 3/3] =?UTF-8?q?fix(main.py):=20=E8=A7=A3=E5=86=B3OpenCV?= =?UTF-8?q?=E4=B8=AD=E6=96=87=E6=A0=87=E6=B3=A8=E4=B9=B1=E7=A0=81+Matplotl?= =?UTF-8?q?ib=E7=BB=98=E5=88=B6=E4=B8=AD=E6=96=87?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/openpilot_model/main.py | 366 +++++++++++++++++++++--------------- 1 file changed, 216 insertions(+), 150 deletions(-) diff --git a/src/openpilot_model/main.py b/src/openpilot_model/main.py index 43b6406a93..d21d10669a 100644 --- a/src/openpilot_model/main.py +++ b/src/openpilot_model/main.py @@ -1,226 +1,292 @@ #!/usr/bin/env python3 -# -*- coding: utf-8 -*- # 声明编码,解决中文注释/输出乱码 +# -*- coding: utf-8 -*- """ -车道线预测程序(优化版) -核心功能:读取MP4视频帧 → 预处理 → 模型推理 → 车道线/路径可视化 -适配环境:Linux虚拟机(Python3 + TensorFlow + OpenCV + Matplotlib) +车道线预测程序(最终版·标注中文正常显示) +核心:用Matplotlib绘制中文标注(替代OpenCV的putText) """ import sys import os +import logging +import argparse +import time import numpy as np import cv2 +import matplotlib import matplotlib.pyplot as plt -from tqdm import tqdm from tensorflow.keras.models import load_model -# ===================== 基础配置(核心!解决中文乱码+路径问题) ===================== -# 项目根目录(绝对路径,适配虚拟机) +# ===================== 环境初始化(核心解决中文显示) ===================== +logging.basicConfig( + level=logging.INFO, + format="%(asctime)s - %(levelname)s - %(message)s", + datefmt="%Y-%m-%d %H:%M:%S" +) +logger = logging.getLogger(__name__) + +# 设置Matplotlib后端 +matplotlib.use('Agg') if os.environ.get('DISPLAY') is None else matplotlib.use('TkAgg') + +# 加载中文字体(仅给Matplotlib用) +def setup_chinese_font(): + font_path = '/usr/share/fonts/truetype/wqy/wqy-microhei.ttc' + if os.path.exists(font_path): + font_prop = matplotlib.font_manager.FontProperties(fname=font_path) + plt.rcParams['font.sans-serif'] = [font_prop.get_name()] + logger.info(f"✅ 中文字体加载成功:{font_path}") + else: + logger.warning("⚠️ 未找到wqy-microhei字体,使用默认英文字体") + plt.rcParams['font.sans-serif'] = ['DejaVu Sans'] + plt.rcParams['axes.unicode_minus'] = False + plt.rcParams['figure.dpi'] = 96 + plt.rcParams['savefig.dpi'] = 100 + +setup_chinese_font() + +# 项目根目录 PROJECT_ROOT = "/home/dacun/nn" sys.path.append(PROJECT_ROOT) -# 解决Matplotlib中文显示乱码(关键优化) -plt.rcParams['font.sans-serif'] = ['DejaVu Sans', 'SimHei', 'WenQuanYi Micro Hei'] -plt.rcParams['axes.unicode_minus'] = False # 解决负号显示问题 +# 依赖检测 +def check_dependencies(): + required_libs = { + 'numpy': np.__version__, + 'cv2': cv2.__version__, + 'matplotlib': matplotlib.__version__, + 'tensorflow': '2.x' + } + for lib, ver in required_libs.items(): + try: + if lib == 'tensorflow': + import tensorflow as tf + assert tf.__version__.startswith('2.'), f"TensorFlow版本需≥2.0,当前:{tf.__version__}" + logger.info(f"✅ 依赖检测通过:{lib} (版本:{ver})") + except (ImportError, AssertionError) as e: + logger.error(f"❌ 依赖缺失/版本错误:{lib} - {e}") + logger.error(f"💡 修复命令:pip install {lib}>={ver.split('.')[0]}") + sys.exit(1) -# 导入项目本地模块 +# 导入项目模块 try: from common.transformations.camera import transform_img, eon_intrinsics from common.transformations.model import medmodel_intrinsics from common.tools.lib.parser import parser + logger.info("✅ 项目模块导入成功") except ImportError as e: - print(f"❌ 导入common模块失败:{e}") - print("💡 请确保common文件夹在项目根目录:/home/dacun/nn/common/") + logger.error(f"❌ 项目模块导入失败:{e}") + logger.error("💡 确认common文件夹路径:/home/dacun/nn/common/") sys.exit(1) -# ===================== 核心函数(规范化+精简冗余) ===================== -def frames_to_tensor(frames): - """ - 将视频帧转换为模型输入张量 - :param frames: 原始视频帧数组 (N, H, W, C) - :return: 归一化后的张量 (N, 6, H//2, W//2) - """ - if len(frames) == 0: +# ===================== 核心函数 ===================== +def frames_to_tensor(frames: np.ndarray) -> np.ndarray: + if frames.size == 0: + logger.warning("输入帧为空,返回空张量") return np.array([]) H = (frames.shape[1] * 2) // 3 W = frames.shape[2] tensor = np.zeros((frames.shape[0], 6, H//2, W//2), dtype=np.float32) - # 张量维度映射(模型输入要求) tensor[:, 0] = frames[:, 0:H:2, 0::2] tensor[:, 1] = frames[:, 1:H:2, 0::2] tensor[:, 2] = frames[:, 0:H:2, 1::2] tensor[:, 3] = frames[:, 1:H:2, 1::2] tensor[:, 4] = frames[:, H:H+H//4].reshape(-1, H//2, W//2) tensor[:, 5] = frames[:, H+H//4:H+H//2].reshape(-1, H//2, W//2) - return tensor / 128.0 - 1.0 # 归一化到[-1, 1] + return tensor / 128.0 - 1.0 -def preprocess_frames(imgs): - """ - 视频帧预处理(适配模型输入格式) - :param imgs: 原始YUV帧列表 - :return: 预处理后的张量 - """ +def preprocess_frame(img: np.ndarray) -> np.ndarray: + try: + return transform_img( + img, + from_intr=eon_intrinsics, + to_intr=medmodel_intrinsics, + yuv=True, + output_size=(512, 256) + ) + except Exception as e: + logger.warning(f"单帧预处理失败:{e},返回空帧") + return np.zeros((384, 512), dtype=np.uint8) + +def preprocess_frames(imgs: list) -> np.ndarray: if not imgs: return np.array([]) - processed = np.zeros((len(imgs), 384, 512), dtype=np.uint8) - # 精准捕获异常,避免通捕导致问题隐藏 - for i, img in enumerate(imgs): - try: - processed[i] = transform_img( - img, - from_intr=eon_intrinsics, - to_intr=medmodel_intrinsics, - yuv=True, - output_size=(512, 256) - ) - except (TypeError, ValueError) as e: - print(f"⚠️ 第{i+1}帧预处理失败:{str(e)},填充空帧") - processed[i] = np.zeros((384, 512), dtype=np.uint8) - return frames_to_tensor(processed) + processed_frames = [preprocess_frame(img) for img in imgs] + processed_frames = np.array(processed_frames) + empty_frames = np.sum(np.all(processed_frames == 0, axis=(1, 2))) + if empty_frames > 0: + logger.warning(f"共{empty_frames}帧预处理失败,已填充空帧") + return frames_to_tensor(processed_frames) -def read_video_frames(video_path, max_frames=10): - """ - 读取视频帧(仅支持MP4),简化函数名更直观 - :param video_path: 视频文件路径 - :param max_frames: 最大读取帧数 - :return: 预处理用YUV帧 + 原始BGR帧 - """ - # 精准校验视频格式 +def read_video(video_path: str, max_frames: int = 10) -> tuple: if not os.path.exists(video_path): raise FileNotFoundError(f"视频文件不存在:{video_path}") if not video_path.lower().endswith('.mp4'): - raise ValueError("仅支持MP4格式视频,请更换文件格式") - + raise ValueError("仅支持MP4格式视频") cap = cv2.VideoCapture(video_path) if not cap.isOpened(): raise RuntimeError(f"无法打开视频(请安装FFmpeg):{video_path}") - - # 降低缓存,减少虚拟机内存占用 + fps = int(cap.get(cv2.CAP_PROP_FPS)) or 10 + width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) + height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) + save_path = os.path.join(PROJECT_ROOT, "lane_pred_result.mp4") + fourcc = cv2.VideoWriter_fourcc(*'mp4v') + video_writer = cv2.VideoWriter(save_path, fourcc, fps, (width, height)) + logger.info(f"✅ 结果视频保存路径:{save_path}") cap.set(cv2.CAP_PROP_BUFFERSIZE, 1) yuv_frames = [] raw_frames = [] - - # 进度条可视化(核心保留优化) - for i in tqdm(range(max_frames), desc="读取视频帧", ncols=80): + for i in range(max_frames): ret, frame = cap.read() if not ret: - tqdm.write(f"⚠️ 视频读取完毕,共读取{i}帧(不足{max_frames}帧)") + logger.info(f"视频读取完毕,共读取{i}帧(目标:{max_frames}帧)") break raw_frames.append(frame) - # BGR转YUV_I420(模型输入要求) yuv = cv2.cvtColor(frame, cv2.COLOR_BGR2YUV_I420) yuv_resized = cv2.resize(yuv, (512, 384), interpolation=cv2.INTER_LINEAR) yuv_frames.append(yuv_resized) - cap.release() - return yuv_frames, raw_frames + return yuv_frames, raw_frames, video_writer -# ===================== 主函数(精简+可视化升级) ===================== -def main(): - # 1. 参数校验(精简且专业) - if len(sys.argv) != 2: - print("🚨 使用错误:缺少视频文件路径") - print("✅ 正确用法:python main.py <视频文件绝对路径>") - print("💡 示例:python main.py /home/dacun/nn/test.mp4") - sys.exit(1) - video_path = sys.argv[1] +def draw_lane_lines(frame: np.ndarray, pred: dict) -> np.ndarray: + """仅绘制车道线(文字标注交给Matplotlib)""" + h, w = frame.shape[:2] + pred_lll = np.interp(pred["lll"][0], (0, 191), (0, h)) + pred_rll = np.interp(pred["rll"][0], (0, 191), (0, h)) + pred_path = np.interp(pred["path"][0], (0, 191), (0, h)) + x_coords = np.linspace(0, w, len(pred_lll)) + frame_copy = frame.copy() + # 左车道线(蓝) + for i in range(len(x_coords)-1): + cv2.line( + frame_copy, + (int(x_coords[i]), int(pred_lll[i])), + (int(x_coords[i+1]), int(pred_lll[i+1])), + (255, 0, 0), 3, cv2.LINE_AA + ) + # 右车道线(红) + for i in range(len(x_coords)-1): + cv2.line( + frame_copy, + (int(x_coords[i]), int(pred_rll[i])), + (int(x_coords[i+1]), int(pred_rll[i+1])), + (0, 0, 255), 3, cv2.LINE_AA + ) + # 预测路径(绿) + for i in range(len(x_coords)-1): + cv2.line( + frame_copy, + (int(x_coords[i]), int(pred_path[i])), + (int(x_coords[i+1]), int(pred_path[i+1])), + (0, 255, 0), 2, cv2.LINE_AA + ) + return cv2.addWeighted(frame_copy, 0.7, frame, 0.3, 0) - # 2. 加载模型(精准路径+异常捕获) +# ===================== 主函数 ===================== +def main(): + parser_arg = argparse.ArgumentParser(description="车道线预测程序(标注中文正常)") + parser_arg.add_argument("video_path", type=str, help="视频文件绝对路径") + parser_arg.add_argument("--max-frames", type=int, default=10, help="最大读取帧数") + parser_arg.add_argument("--save-result", action="store_true", default=True, help="保存结果视频") + args = parser_arg.parse_args() + + check_dependencies() + + # 加载模型 model_path = os.path.join(PROJECT_ROOT, "models/supercombo.h5") if not os.path.exists(model_path): - raise FileNotFoundError(f"模型文件不存在:{model_path}(请放入models目录)") - + logger.error(f"模型文件不存在:{model_path}") + sys.exit(1) try: - print(f"📌 加载模型:{model_path}") + logger.info(f"开始加载模型:{model_path}") + start_time = time.time() model = load_model(model_path, compile=False) - except (IOError, ValueError) as e: - print(f"❌ 模型加载失败:{str(e)}") + logger.info(f"✅ 模型加载完成,耗时{round(time.time()-start_time,2)}秒") + except Exception as e: + logger.error(f"❌ 模型加载失败:{e}") sys.exit(1) - - # 3. 读取+预处理视频帧 + + # 读取视频 try: - yuv_frames, raw_frames = read_video_frames(video_path) - if not yuv_frames: - raise RuntimeError("未读取到有效视频帧") - frame_tensor = preprocess_frames(yuv_frames) - if frame_tensor.size == 0: - raise RuntimeError("帧预处理后无有效数据") + yuv_frames, raw_frames, video_writer = read_video(args.video_path, args.max_frames) + if not raw_frames: + logger.error("未读取到有效视频帧") + sys.exit(1) + logger.info(f"✅ 视频读取完成,共{len(raw_frames)}帧") except Exception as e: - print(f"❌ 视频处理失败:{str(e)}") + logger.error(f"❌ 视频读取失败:{e}") sys.exit(1) - - # 4. 模型推理初始化 - state = np.zeros((1, 512)) # 模型状态初始化 - desire = np.zeros((1, 8)) # 行驶意图初始化 + + # 预处理 + frame_tensor = preprocess_frames(yuv_frames) + if frame_tensor.size == 0: + logger.error("帧预处理后无有效数据") + sys.exit(1) + + # 推理初始化 + state = np.zeros((1, 512)) + desire = np.zeros((1, 8)) total_frames = len(frame_tensor) - 1 - - # 5. 可视化升级(解决乱码+效果优化) - plt.ion() # 交互式模式 - fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5)) # 分屏显示:原始帧+预测结果 - fig.suptitle("车道线预测结果", fontsize=14, fontweight='bold') - - # 子图1:原始视频帧 - ax1.set_title("原始视频帧", fontsize=12) - ax1.axis('off') # 关闭坐标轴,更清晰 - img_display = ax1.imshow(cv2.cvtColor(raw_frames[0], cv2.COLOR_BGR2RGB)) - - # 子图2:车道线预测(优化线条+标注) - ax2.set_title("车道线/路径预测", fontsize=12) - ax2.set_xlabel("横向像素", fontsize=10) - ax2.set_ylabel("纵向像素", fontsize=10) - ax2.set_ylim(0, 191) - ax2.invert_xaxis() # 匹配驾驶视角(左/右对齐) - ax2.grid(alpha=0.2, linestyle='--') # 轻量化网格 - - # 初始化预测线条(颜色标准化+标签清晰) - left_line, = ax2.plot([], [], 'b-', linewidth=2.5, label='左车道线') - right_line, = ax2.plot([], [], 'r-', linewidth=2.5, label='右车道线') - path_line, = ax2.plot([], [], 'g-', linewidth=2, label='预测路径') - ax2.legend(loc='lower left', fontsize=9) # 图例位置优化 - - # 6. 逐帧推理+可视化更新 - print(f"\n🚀 开始推理(共{total_frames}帧,按Q键退出)") - try: - for i in range(total_frames): - # 模型推理(核心逻辑无改动) + logger.info(f"开始推理,共{total_frames}帧(按Q键退出)") + + # 可视化(用Matplotlib添加中文标注) + plt.ion() + fig, ax = plt.subplots(figsize=(10, 6)) + fig.suptitle("车道线预测结果(叠加可视化)", fontsize=14, fontweight='bold') + # 添加中文图例标注(Matplotlib支持中文) + ax.text( + 0.02, 0.95, + "左车道线(蓝) | 右车道线(红) | 预测路径(绿)", + transform=ax.transAxes, + fontsize=10, + color='white', + bbox=dict(facecolor='black', alpha=0.5) + ) + ax.set_axis_off() + img_display = ax.imshow(cv2.cvtColor(raw_frames[0], cv2.COLOR_BGR2RGB)) + + for i in range(total_frames): + try: + # 推理 input_tensor = np.vstack(frame_tensor[i:i+2])[None] outputs = model.predict([input_tensor, desire, state], verbose=0) pred_result = parser(outputs) state = outputs[-1] - - # 更新预测线条(对齐维度) - left_line.set_data(pred_result["lll"][0], range(192)) - right_line.set_data(pred_result["rll"][0], range(192)) - path_line.set_data(pred_result["path"][0], range(192)) - - # 更新原始帧显示 - if i < len(raw_frames): - img_display.set_data(cv2.cvtColor(raw_frames[i], cv2.COLOR_BGR2RGB)) - - # 刷新画布 + + # 绘制车道线 + result_frame = draw_lane_lines(raw_frames[i], pred_result) + + # 更新显示 + img_display.set_data(cv2.cvtColor(result_frame, cv2.COLOR_BGR2RGB)) fig.canvas.draw() fig.canvas.flush_events() - - # 键盘退出(仅保留Q键,删除冗余自动退出) - if cv2.waitKey(50) & 0xFF == ord('q'): - print("🛑 用户按Q键退出推理") + + # 保存结果 + if args.save_result: + video_writer.write(result_frame) + + # 退出 + if cv2.waitKey(30) & 0xFF == ord('q'): + logger.info("用户按Q键退出") break - - print(f"✅ 完成第{i+1}/{total_frames}帧推理") - - finally: - # 资源释放(彻底+规范) - print("\n🧹 释放资源中...") - plt.ioff() - plt.close(fig) - cv2.destroyAllWindows() - # 强制清除CV2残留 - cv2.waitKey(1) - print("🎉 程序正常结束") + + logger.info(f"✅ 完成第{i+1}/{total_frames}帧推理") + + except Exception as e: + logger.warning(f"⚠️ 第{i+1}帧推理失败:{e},跳过") + continue + + # 资源释放 + plt.ioff() + plt.close(fig) + video_writer.release() + cv2.destroyAllWindows() + cv2.waitKey(1) + + # 成果提示 + logger.info("\n🎉 程序执行完成!") + logger.info(f"📁 结果视频:{os.path.join(PROJECT_ROOT, 'lane_pred_result.mp4')}") if __name__ == "__main__": - # 全局异常捕获(更专业) try: main() except Exception as e: - print(f"\n❌ 程序异常终止:{str(e)}") + logger.error(f"\n❌ 程序异常终止:{e}") sys.exit(1)