服务异常(TypeError: object of type 'NoneType' has no len())

[hebutliu2010]

[2023/12/06 09:48:47] ppocr DEBUG: dt_boxes num : 1, elapsed : 1.3040359020233154
[2023/12/06 09:48:47] ppocr DEBUG: cls num : 1, elapsed : 0.038794517517089844
[2023/12/06 09:48:47] ppocr DEBUG: rec_res num : 1, elapsed : 0.24462223052978516
Traceback (most recent call last):
File "/bin/python3.9/bin/wesubtitle", line 33, in
sys.exit(load_entry_point('wesubtitle==0.0.1', 'console_scripts', 'wesubtitle')())
File "/bin/python3.9/lib/python3.9/site-packages/wesubtitle-0.0.1-py3.9.egg/wesubtitle/main.py", line 129, in main
detected, box, content = detect_subtitle_area(ocr_results, h, w)
File "/bin/python3.9/lib/python3.9/site-packages/wesubtitle-0.0.1-py3.9.egg/wesubtitle/main.py", line 28, in detect_subtitle_area
while idx < len(ocr_results):

Uploading test.mp4…

[L1Star03]

我也遇到了类似的问题，请问您解决了吗

[L1Star03]

问题已解决，可以尝试通过修改库中main.py中的函数解决

def detect_subtitle_area(ocr_results, h, w):
    '''
    Args:
        w(int): width of the input video
        h(int): height of the input video
    '''
    # print(ocr_results) # for debug add by L1_Sta2
    ocr_results = ocr_results[0]  # 0, the first image result
    
    # Merge horizon text areas
    idx = 0
    candidates = []
    if not ocr_results:
        ocr_results = []
        
    while idx < len(ocr_results):
        boxes, text = ocr_results[idx]
        # We assume the subtitle is at bottom of the video
        if boxes[0][1] < h * 0.75:
            idx += 1
            continue
        idx += 1
        con_boxes = copy.deepcopy(boxes)
        con_text = text[0]
        while idx < len(ocr_results):
            n_boxes, n_text = ocr_results[idx]
            if abs(n_boxes[0][1] - boxes[0][1]) < h * 0.01 and \
               abs(n_boxes[3][1] - boxes[3][1]) < h * 0.01:
                con_boxes[1] = n_boxes[1]
                con_boxes[2] = n_boxes[2]
                con_text = con_text + ' ' + n_text[0]
                idx += 1
            else:
                break
        candidates.append((con_boxes, con_text))
    # TODO(Binbin Zhang): Only support horion center subtitle
    if len(candidates) > 0:
        sub_boxes, subtitle = candidates[-1]
        # offset is less than 10%
        if (sub_boxes[0][0] + sub_boxes[1][0]) / w > 0.70: # 这里降低了要求
            return True, box2int(sub_boxes), subtitle
    return False, None, None

原因可能是视频的前几帧或几秒钟没有出现在合适位置的字幕，导致ocr_result返回None值，而在上述函数中没有对None的处理逻辑，直接使用了len方法，所以这里修改如果为None设置为空列表，在不改变后续操作的情况下跳过循环。
感觉也可以考虑在进行ocr之前先进行裁切。

[henry-tujia]

You save my DAY!!