deep-collaboration-network

KO | EN

gitlite — search

#javascript #python #ai #llm #go #android #machine-learning #docker #java #kubernetes #vue #awesome-list #pytorch #openai

deep-collaboration-network

(★ 15)

Deep Collaboration Network in pytorch

deep-collaboration-network Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

Deep Collaboration Network
============================

This is an implementation of Deep Collaboration Network (DCNet) in `pytorch` from [Multi-Task Learning by Deep Collaboration and Application in Facial Landmark Detection](https://arxiv.org/abs/1711.00111) by Trottier, et al. (2017).

![](https://raw.githubusercontent.com/ltrottier/deep-collaboration-network/master/collaborative-block.png)

## Requirements

1. [pytorch](http://pytorch.org/)
2. torchvision==0.1.8

## Update

The collaborative block now uses an additional ReLU after the sum between the input (coming from the identity skip connection) and the output of the task aggregation. We obtained better performance with it.

## Example

Here is an example on how to use `DCNet` and train it with `MultiTaskCriterion`:

```python
from dcnet import DeepCollaborationNetwork, MultiTaskCriterion
from torch.autograd import Variable
import numpy as np
import torch
import torch.optim as optim
import torch.nn as nn

# define (input, targets) pair
dims = [10, 5]
bs = 32
input = Variable(torch.randn(bs, 3, 112, 112))
targets = []
for i, dim in enumerate(dims):
    target = np.random.randint(0, dim, (bs, ))
    target = torch.from_numpy(target)
    target = Variable(target)
    targets.append(target)

# create training criterion
criterions = [nn.CrossEntropyLoss() for _ in dims]
weights = [1] * len(dims)
criterion = MultiTaskCriterion(criterions, weights)

# create network
net = DeepCollaborationNetwork('resnet18', dims, pretrained=True)

# optimize
optimizer = optim.SGD(net.parameters(), lr=0.1, momentum=0.9, nesterov=True)
optimizer.zero_grad()
output = net(input)
loss = criterion(output, targets)
loss.backward()
optimizer.step()

```

deep-collaboration-network

deep-collaboration-network Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

deep-collaboration-network

deep-collaboration-network Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README