Skip to content

Latest commit

 

History

History
42 lines (27 loc) · 1.33 KB

File metadata and controls

42 lines (27 loc) · 1.33 KB

Weakly-supervised-object-localization

simple weakly dectector

we rewrite two version:
keras + tensorflow2
pytorch

dataset

cifar10, it can be loaded by keras or pytorch with one line

author's paper

Learning Deep Features for Discriminative Localization

code

using the InceptionV3 to extract conv_features
after 3 iterations,training accuracy is 99.87,testing accuracy is 95.68

Step:
(1):run train.py
(2):run draw_detector.py

also,some tricks are learned from:https://github.com/jazzsaxmafia/Weakly_detector

Results show

Here we just show some samples...

Sample 1: sample and hotmat 1 combine 1

Sample 2: sample and hotmat 1 combine 1

Sample 3: sample and hotmat 1 combine 1