yingzhangdut / deep-mutual-learning Goto Github PK

Hello, I am a beginner of deep learning.I read the code you wrote about the market1501 dataset.
I tried the cifar100 dataset myself, but it didn't work. Can you share the code about the cifar100 dataset, thank you! Sincerely need your help.

Hello, firstly, thanks your demo. When using pre-trained model, we need modify the variable name of official model in order to keep in line with our variable name in our model. My method is that I read official model ,modify variable name and save it in python file. But the operation is a little tedious! Could you tell me what you have done?

I have trained the model on Market 1501. Now i want to use it on my own dataset for fine tuning and evaluation. How can i do that?

I tested Resnet-32 (both Independent and DML) on Cifar100 dataset, but I didn't get significant improvement using DML, was my experimental settings incorrect?

I followed your experimental settings, which is:

• learning rate
  • 0.1 (epoch < 60)
  • 0.01 (epoch < 120)
  • 0.001 (epoch < 180)
  • 0.0001 (otherwise)
• momentum 0.9
• batch-size 64
• epoch 200
• weight-decay 5e-4
• using SGD
  • set nesterov=True

my result of independent resnet-32 is (average of 6 results) 70.98% , and DML of resnet-32 and resnet-32 is 70.86% / 70.90% .

Hello ! Is there a code for cifar datasets of deep mutual learning? I don't know much about market-1501, I want to learn the code of cifar dataset.Can you share code for cifar datasets of deep mutual learning ?Thanks!

Whether the two stream networks of this model are trained at the same time?can keras realize it ?

when I run train code，I have met this problem，how to solve it ？？
ERROR:tensorflow:Exception in QueueRunner: assertion failed: [Unable to decode bytes as JPEG, PNG, GIF, or BMP]
[[Node: case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/Assert_1/Assert = Assert[T=[DT_STRING], summarize=3, _device="/job:localhost/replica:0/task:0/cpu:0"](case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/is_bmp, case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/Assert_1/Assert/data_0)]]
Exception in thread Thread-6:
Traceback (most recent call last):
File "/home/wangliu/anaconda3/lib/python3.6/threading.py", line 916, in _bootstrap_inner
self.run()
File "/home/wangliu/anaconda3/lib/python3.6/threading.py", line 864, in run
self._target(*self._args, **self._kwargs)
File "/home/wangliu/anaconda3/lib/python3.6/site-packages/tensorflow/python/training/queue_runner_impl.py", line 238, in _run
enqueue_callable()
File "/home/wangliu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1235, in _single_operation_run
target_list_as_strings, status, None)
File "/home/wangliu/anaconda3/lib/python3.6/contextlib.py", line 88, in exit
next(self.gen)
File "/home/wangliu/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py", line 466, in raise_exception_on_not_ok_status
pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors_impl.InvalidArgumentError: assertion failed: [Unable to decode bytes as JPEG, PNG, GIF, or BMP]
[[Node: case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/Assert_1/Assert = Assert[T=[DT_STRING], summarize=3, _device="/job:localhost/replica:0/task:0/cpu:0"](case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/is_bmp, case/If_0/decode_image/cond_jpeg/cond_png/cond_gif/Assert_1/Assert/data_0)]]

Hello everyone:
I have a quetsion about how to evaluate after I done the "sh scripts/evaludate_dml_mobilenet_on_market.sh" and I got this .

When I open Market_1501_evaluation.m , I don't know how to run this code???
Here is the question, I got six files that is. Could you tell me more details about how to do that??

Please release the dataset in baidu pan.

I cant achieve the 73.65% of mobilenet in the paper and only 67%.
My setting is as follows:
lr = 0.1
epoch = 200
lr-divide =0.1 in [60,120,180]
weigh_decay = 0.0005
momontum = 0.9

there are always some bugs。

I am getting a 404 when I am trying to download the dataset over this link:
http://www.liangzheng.com.cn/Project/project_reid.html
Can you provide me with another link?
Thanks

2. Restore the pretrained checkpoint to "MobileNetV1".
3. Assign all the variables under "MobileNetV1" to "net1";

Originally posted by @YingZhangDUT in #4 (comment)

Hi, Thank you for this excellent code. I am new to the cifar100 dataset. The result confused me that cifar100 acc is much higher in other repos:
https://github.com/weiaicunzai/pytorch-cifar100
resnet34 got 23.24 error rate
and much higher in self distillation
https://github.com/luanyunteng/pytorch-be-your-own-teacher
Thanks for your time.

Hi, I am confused with the experiment on Market-1501 in section 3.3.

In this experiment, net1 = MobileNet, net2 = MobileNet, and they are both pre-trained model in ImageNet_Pretrain = yes (last line) settings. Thus, net1 is exactly the same with net2, right?

So, the prediction of net1 and net2 should be exactly the same because MobileNet does not have Dropout. Thus, p1 = p2, and Dkl = 0, and Loss1 = Loss2, right?

Thus, after training, net1 should be exactly the same with net2, right?

So, DML in this settings has Dkl = 0, and this has little difference with training single MobileNet. Why it has a huge performance improvement (rank-1 87.73 vs 83.94)?

The paper also says "Models Θ1 and Θ2 to different initial conditions". How does this relate to pre-trained models?

I'm a little confused.
In your paper, the 3rd step of Algorithm 1 is "Update the predictions p1 && p2 ..."
but the implementation here doesn't update the predicts.

when i run $ sh scripts/format_and_convert_market.sh,

Thanks!