makecent

Segmentation fault (core dumped) when few video frames

7

My denseflow running always abort with error `Segmentation fault (core dumped)` at the middle of task. I investigated the error information, and found that it always happens when the number...

Maybe we should support wild backbones/models ?

8

**Describe the feature** Add a `wild_backbones` folders that stores the backbones directly copied from the original papers/repositories. They are just about to work and could be maintained by any contributors....

## Motivation To use slowfast models with mmaction2. #1711 ## Modification Adding a mini `slowfast` package under `models/contrib`. It's copied from the [SlowFast](https://github.com/facebookresearch/SlowFast/tree/main/slowfast) and pruned to only keep neccesary components...

This PR takes the first step of #1711. Just a demo. **Motivation** If we allow wild backbones, the latest models (with code released) can be quickly adopted to mmation2. For...

label smoothing should NOT only work under multi-class settings

2

Labeling smoothing is a widely used regularation method in action recognition tasks. However, currenly `label_smooth_eps` only works when the `multi-class` is toggled on (which default `False`): https://github.com/open-mmlab/mmaction2/blob/6d6685632f28344e98cf34a14d1226cd6c008391/mmaction/models/heads/base.py#L106-L108 I suggest to...

"evaluation" and "eval_config" are confusing

4

I found that config files always contain a "evaluation" dict and a "eval_config". The `evaluation` is used in the **validation stage** : https://github.com/open-mmlab/mmaction2/blob/99a0e0a7d4cceb4568dab765c53970e5c83dc7e9/mmaction/apis/train.py#L179-L180 while the `eval_config` is used in the...

Correct Resize pipeline

9

Fix #209

height and width in Resize pipeline are misplaced

2

**Reproduction** ```python from mmgen.datasets.pipelines import Resize, LoadImageFromFile results = {'real_img_path': 'path-to-img/0001.jpg'} t = LoadImageFromFile(key='real_img', io_backend='disk')(results) tt = Resize(keys=['real_img'], scale=(128, 32), keep_ratio=False)(t) ``` **Bug fix** I think it's because mmcv.imresize uses...

Please compare with regression network

3

According to the design of your CORAL framework, it is clear that the output in the penultimate layer, which has only 1 node, is proportional to the age of the...

Feature extraction for THUMOS14 is strange

4

According to the "Implementation Details" part in Section 4 of the original paper, you use the model pre-trained on the training set of ActivityNet-1.3 as the features extractor. And I...