open-mmlab/mmaction2

"prec@top5": 0.2,为什么一直都是0.2，不应该靠近1？？还有日志中只有recall@top3、prec@top3、recall@prec@top5。我该怎么做才能获得 ACC（top1\top5） [Bug]

Open

#2624 opened on Aug 3, 2023

View on GitHub
 (1 comment) (0 reactions) (1 assignee)Python (3,648 stars) (1,137 forks)batch import
help wanted

Description

Branch

main branch (1.x version, such as v1.0.0, or dev-1.x branch)

Prerequisite

Environment

cuda11.3 py3.8

Describe the bug

第一个问题：日志中只有recall@top3、prec@top3、recall@prec@top5。我该怎么做才能获得 ACC（top1\top5） 第二个问题：{"mode": "train", "epoch": 5, "iter": 460, "lr": 0.0084, "memory": 3772, "data_time": 1.31733, "recall@thr=0.5": 0.66107, "prec@thr=0.5": 0.64927, "recall@top1": 0.7646, "prec@top1": 0.7646, "recall@top5": 1.0, "prec@top5": 0.2, "loss_action_cls": 0.23607, "loss": 0.23607, "grad_norm": 0.98712, "time": 1.8695} "prec@top5": 0.2,为什么一直都是0.2，不应该靠近1？？

Reproduces the problem - code sample

正常训练，就是训练结果不理想

Reproduces the problem - command or script

正常训练，就是训练结果不理想

Reproduces the problem - error message

正常训练，就是训练结果不理想

Additional information

正常训练，就是训练结果不理想。在标注框时官方提供是fastrcnn，我用的是2023最新的grouonddino并且成功制作了ava数据集，model = dict( type='FastRCNN', backbone=dict( type='ResNet3dSlowFast', 难道训练也要后期配置也要用grouonddino？？

Contributor guide

Tech stack
pythonpytorch
Domain
machine learningai
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
PythonPyTorchmmaction2
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
25
Research direction
Investigate the evaluation configuration in mmaction2 to understand why prec@top5 is fixed at 0.2 and how to add top1/top5 accuracy. Check the configuration file for the dataset and evaluation metrics. Look into the code for metric computation in the `mmaction/evaluation/` directory to see if there is a bug or default setting that causes prec@top5 to be constant. Also consider the model output format perhaps the model is not outputting top5 properly.