Search for resnet implementation in tensor stream

Question

Are there any resnet implementers in the tensor stream? I came across a few (e.g. https://github.com/ry/tensorflow-resnet , https://github.com/xuyuwei/resnet-tf ), but there are some errors in these implementations (e.g. see " Problems "on the corresponding github page). I am looking to train imagenet using resnet, and am looking for implementations of tensor flow.

Answer 1

There are several (50/101/152) in tensorflow: models / slim .

The laptop example shows how to get a prepared inception start, res-net is probably no different.

Answer 2

I implemented a version of Reset cifar10 with tensor flow. Verification errors ResNet-32, ResNet-56 and ResNet-110 are 6.7%, 6.5% and 6.2%, respectively. (You can easily change the number of layers as hyperparameters.)

I tried to be friendly with the new ResNet fan and it was simple to write. You can run the cifar10_train.py file directly without any downloads.

https://github.com/wenxinxu/resnet_in_tensorflow

Answer 3

Here is an implementation I created using the TensorFlow GPU in Docker, using the full ImageNet dataset trained on 2x NVIDIA RTX 2080 Tis.

https://blog.exxactcorp.com/deep-learning-with-tensorflow-training-resnet-50-from-scratch-using-the-imagenet-dataset/

It is mainly based on the official implementation of TensorFlow, with some tips and tricks that I found along the way. This also applies to launching and interacting with the TF dock environment.

Let me know if you have any problems or questions!

James

Answer 4

I implemented Resnet using ronnie.ai and keras. Both tools are great. Although Ronnie is easier from scratch.