regainOWO

Use TCP protocol and RTSP Stream to read frame, and convert it to numpy array, but i don't want to save the frame, just read

4

i have check out the demo [here](https://kkroening.github.io/ffmpeg-python/#ffmpeg.input), the specificd location is in title (Process video frame-by-frame using numpy) . because i want to just read and convert it to numpy...

batch-size: 32 device: Tesla V100 32GB model: yolox s 自己训练的旋转框模型 num_class: 11 765张图片用于验证，结果如下  然后取765张图片中的99张，进行测试，结果如下  **可以发现测试结果的推理时间相差较大，测试期间并无其他程序占用测试device的情况**

module 'numpy' has no attribute 'int' [diffs = ann.get('diffs', np.zeros((gt_bboxes.shape[0], ), dtype=np.int))](https://github.com/DDGRCF/YOLOX_OBB/blob/986dbee50c7a0ccbeb4c7db94d672db8a2a7e3a2/yolox/data/datasets/dota.py#L390) maybe you can change dtype=int module 'numpy' has no attribute 'bool' [np.zeros(gt_bboxes.shape[0], dtype=np.bool)](https://github.com/DDGRCF/YOLOX_OBB/blob/986dbee50c7a0ccbeb4c7db94d672db8a2a7e3a2/yolox/evaluators/dota_eval.py#L228) numpy only have and `bool_`...

SCD模型在SECOND数据集上训练，验证时，出现了NaN

7

环境: * python: 3.9.0 * torch: 1.10.1 * mmcv: 2.1.0 * mmsegmentation: 1.2.2 * opencd: 1.1.0 scd_upernet在SECOND数据集上训练，但在验证的时候分数为NaN。 log日志如下： ```bash 2024/04/26 10:57:47 - mmengine - INFO - ------------------------------------------------------------ System environment: sys.platform:...

SelectiveScanEasy forward result have Nan

1

this project is amazing! i change the SelectiveScanOflex to SelectiveScanEasy for cpu inference, but the result have Nan,  i found it cause by `tmp_dtBus_div_rAts = (dtBus / rAts)`, the...

I have a lot of `.las` point cloud files and I want to input them all into LaserMix for training, I have a few questions below: 1. Its XYZ is...