cir7

@WP-CV hi, our implement use nn.MultiheadAttention. when q, k, v are same tensor(ie. query_t), it will perform a self-attention projection first, and get the projected tensor Q, K, V. then...

The error message indicates that there are too many values for each annotation line. Please check and make sure that your annotation file has the same format as follows: ```...

could you paste part of your annotation here?

There are whitespace characters in your video file name, and we default to use it as a split delimiter, which results in more values than expected. I suggest removing the...

please make sure the file exist: /mmaction2/mmaction/datasets/classvideo/train/play_phone(2).mp4, maybe you forget to rename video files?

Hi, @sunbaby01, you can modify the `feat_dim` in model config to yours, here is a [tutorial](https://mmaction2.readthedocs.io/en/latest/user_guides/config.html#config-system-for-action-localization) about action localization config.

Hi, @Wrc0217 , you can refer to the [tutorial](https://mmaction2.readthedocs.io/en/latest/datasetzoo.html#activitynet) about prepare activitynet feature to extract your feature, and follow the prepare dataset [tutorial](https://mmaction2.readthedocs.io/en/latest/user_guides/prepare_dataset.html#temporal-action-localization) to prepare the annotations.

This script is outdated and is for version 0.x. We will try to update it soon.

Hi, @Yu-0920 1. `LoadImageFromFile` has been removed in mmdet, you could modify `mmdet.LoadImageFromFile` to `LoadImageFromFile` in `demo/demo_configs/faster-rcnn_r50-caffe_fpn_ms-1x_coco-person.py`. 2. you can load the pickle file with `mmengine.load` and concatenate them as...

We have fixed the script in #2402 , you could try it, and we will merge it into the main branch soon.