About Figure 10 in the paper (Number of unique detectors vs Training iteration)

CSAILVision / NetDissect

Network Dissection http://netdissect.csail.mit.edu for quantifying interpretability of deep CNNs.

MIT License

447 stars 115 forks source link

About Figure 10 in the paper (Number of unique detectors vs Training iteration) #3

Closed Jiangfeng-Xiong closed 7 years ago

Jiangfeng-Xiong commented 7 years ago

Hi, First of all, thank you for your amazing work. How can I get the figure below in the paper? Is it available in this repository?

zhoubolei commented 7 years ago

You need to save the snapshots at each epoch during the training, then run netdissect on each one of them. After that you could plot the figure like this.