1) Implement on AlexNet architecture (IN EAGER MODE)
attention on Conv4, Conv5 layers (Figure 2 of paper)
2) Report Table 1 and Table 2 results (Note that baseline ALexnet results will also have to be collated)
3) Script that visualizes the attention map like in Fig. 3/4
4) Weakly supervised semantic segmentation -- Fig 9 of paper
Due Date Nov4 11:59 PM IST.
Code has to be submitted. Results have to be submitted
Table 1 and 2 rows will be "baseline Alexnet" and "alexnet-att2" (alexnet with attention on conv5 and conv3)"
Fig 9 results will be alexnet-att2 and bonus alexnet-GAP
https://openreview.net/forum?id=HyzbhfWRW
1) Implement on AlexNet architecture (IN EAGER MODE) attention on Conv4, Conv5 layers (Figure 2 of paper)
2) Report Table 1 and Table 2 results (Note that baseline ALexnet results will also have to be collated)
3) Script that visualizes the attention map like in Fig. 3/4
4) Weakly supervised semantic segmentation -- Fig 9 of paper
Due Date Nov4 11:59 PM IST.
Code has to be submitted. Results have to be submitted Table 1 and 2 rows will be "baseline Alexnet" and "alexnet-att2" (alexnet with attention on conv5 and conv3)" Fig 9 results will be alexnet-att2 and bonus alexnet-GAP
tx