paddleOCR在Linux下自动中断程序

[Boom-Wang]

🔎 Search before asking

• [X] I have searched the PaddleOCR Docs and found no similar bug report.
• [X] I have searched the PaddleOCR Issues and found no similar bug report.
• [X] I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

碰到了一个问题：代码在本地Windows环境下运行的时候可以正常运行，并且在测试了一个视频后可以继续测试下一个，不会自动中断。

但是当我把代码放到Linux环境中，测试完一个视频后会自动中断，需要我通过手动启动代码的方式才能继续测试下一个视频

这个问题是当我将LPRNet识别车牌模型换成PaddleOCR模型之后才出现的，之前我用LPRNet来进行车牌识别没有这个问题，但是由于LPRNet识别车牌效果很差，所以换成效果更好的PaddleOCR。

Windows环境下的cuda为11.8，Linux为11.3

🏃‍♂️ Environment (运行环境)

OS Ubuntu 18 Python 3.8 PaddleOCR 2.8.1 CUDA 11.3

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

def recognize_plate_single_image(plate_img): ocr = get_ocr() result = ocr.ocr(plate_img, det=False, cls=True) if result: text = result[0][0][0] confidence = result[0][0][1] return text, confidence else: return "未检测到", 0

@app.route('/car/recognize', methods=['POST']) def recognize_car():

model = get_model()
data = request.get_json()
video_url = data.get('videoUrl')
final_results = []
for track_id, data in vehicle_detections.items():
    carType = Counter(data["carTypes"]).most_common(1)[0][0]
    
    license_plate_groups = {}
    for idx, (num, conf) in enumerate(zip(data["carNums"], data["carNumConfs"])):
        if num not in license_plate_groups:
            license_plate_groups[num] = {'confs': [conf], 'indices': [idx]}
        else:
            license_plate_groups[num]['confs'].append(conf)
            license_plate_groups[num]['indices'].append(idx)

    valid_license_plates = []
    for num, info in license_plate_groups.items():
        confs = info['confs']
        min_conf = min(confs)
        max_conf = max(confs)
        # 判断置信度差异是否小于 0.1
        if max_conf >= 0.95:
            valid_license_plates.append((num, len(confs), info['indices']))
    
    if valid_license_plates:
        # 选择出现次数最多的车牌号
        valid_license_plates.sort(key=lambda x: x[1], reverse=True)
        carNum = valid_license_plates[0][0]
        selected_indices = valid_license_plates[0][2]
        
        # 根据选定的车牌号，获取对应的车牌颜色
        carNumColors_filtered = [data["carNumColors"][i] for i in selected_indices]
        if carNumColors_filtered:
            carNumColor = Counter(carNumColors_filtered).most_common(1)[0][0]
        else:
            carNumColor = "unknown"
    else:
        carNum = "未检测到"
        carNumColor = "unknown"
    
    # 车辆颜色直接统计出现次数最多的
    carColor = Counter(data["carColors"]).most_common(1)[0][0]
    
    result = {
        "carType": carType,
        "carNum": carNum,
        "carColor": carColor,
        "carNumColor": carNumColor,
    }
    final_results.append(result)

vehicle_frames.clear()
vehicle_positions.clear()
return jsonify(final_results)