Pretrain-model输出的cls_score_all、event_times含义

[RyanYChen]

您好，我下载pretrain-model后，使用以下命令，跑10s视频的inference
python3.7 -B -m paddle.distributed.launch --gpus='0' --log_dir=$LOG_DIR main.py --test -c data/soccernet_inference/soccernet_pptimesformer_k400_one_file_inference.yaml -w $WEIGHT_FILE -o inference_dir=$INFERENCE_DIR -o DATASET.test.file_path=$VIDEO_CONFIG

在结果文件中， cls_score_all.shape: [5, 1, 18], event_times.shape: [5, 1, 18]， 请问下这里18维对应的label顺序是什么，以及event_times代表什么含义？

[paulxinzhou]

cls_score_all 和 event_times暂时不用管，当时是为了end 2 end训练预留的，但是后来没来的及实现，所以现在暂时没什么用，feature key下面的数才有用。

[realsazzad]

Hello @paulxinzhou , What these 'features' data mean?
Sorry for asking but really getting a hard time to understand the inference result.

[paulxinzhou]

@realsazzad I added some explanation here. https://github.com/baidu-research/Soccernet-features/blob/develop/README.md#understanding-the-result-features

Let me know if there's still confusion.