# Video-Action-Transformer-Network-Pytorch-
Pytorch and Tensorflow Implementation of the paper Video Action Transformer Network
*Rohit Girdhar, Joao Carreira, Carl Doersch, Andrew Zisserman*
Retasked Video transformer (uses resnet as base)
**transformer_v1.py** is more like real transformer, **transformer.py** more true to what paper advertises
Usage :
```
from transformer_v1 import Semi_Transformer
model = Semi_Transformer(num_classes=num_classes , num_frames = max_seq_len)
outputs, features = model(imgs) # outputs is the classification layer output (do cross entropy loss)
#features are used as video embedding
##################### or ###################
from transformer_v2 import Semi_Transformer
model = Semi_Transformer(num_classes=625 , seq_len = max_seq_len)
```
In case you find any discrepency, please raise an issue. If any one was able to reproduce the paper results kindly help me with this [issue](https://github.com/ppriyank/Video-Action-Transformer-Network-Pytorch-/issues/2). If possible please meantion the changes needs to be further added.
没有合适的资源?快使用搜索试试~ 我知道了~
Video-Action-Transformer-Network-Pytorch-:视频行动变压器网络的实现
共5个文件
py:4个
md:1个
需积分: 50 8 下载量 111 浏览量
2021-05-01
17:31:59
上传
评论
收藏 9KB ZIP 举报
温馨提示
视频动作变压器网络火炬 视频行动变压器网络的Pytorch和Tensorflow实现Rohit Girdhar,Joao Carreira,Carl Doersch,Andrew Zisserman 重新定义视频变压器(用途为RESNET基地)transformer_v1.py更像是真正的变压器,transformer.py更真实到什么纸通告用法: from transformer_v1 import Semi_Transformer model = Semi_Transformer(num_classes=num_classes , num_frames = max_seq_len) outputs, features = model(imgs) # outputs is the classification layer output (do cross entropy loss)
资源详情
资源评论
资源推荐
收起资源包目录
Video-Action-Transformer-Network-Pytorch--master.zip (5个子文件)
Video-Action-Transformer-Network-Pytorch--master
README.md 1KB
tools.py 3KB
transformer_v2.py 7KB
transformer_keras_tensorflow.py 8KB
transformer_v3.py 8KB
共 5 条
- 1
笨猫猪
- 粉丝: 30
- 资源: 4732
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0