See full course on Object Detection: ruclips.net/p/PL1GQaVhO4f_jLxOokW7CS5kY_J1t1T17S If you found this tutorial useful, please share with your friends(WhatsApp/iMessage/Messenger/WeChat/Line/KaTalk/Telegram) and on Social(LinkedIn/Quora/Reddit), Tag @cogneethi on twitter.com Let me know your feedback @ cogneethi.com/contact
Firstly. Awesome video series. I am forwarding to lot of my friends. Correct me if I am wrong. When we do Anchor Box processing in the RPN, we also slide 1 pixel at a time (over the CNN feature maps) and evaluate 9 boxes, right ? So, that's also kind of feature pyramid. But here, you are commenting that such feature pyramid is a time consuming idea !!
Avisek, I think my usage of the term 'Feture Pyramid' here is probably not correct. Here, what I meant was, if we had used Feature Pyramid, then we had to do ROI pooling using 9 different sliding window sizes at each pixel of Feature Map. Instead, by using Anchor Boxes, we do ROI pooling using only 1 square sliding window at each pixel of FM and get 9 different predictions. So, that way we will be saving processing time. Perhaps, it is more clear after watching next video 'C.8.1' and also 'C.8.6 Quirks...' Let me know if I need to elaborate. Let me know if something is not clear or if I have made any mistakes too. Btw, thank you for sharing with your friends!
See full course on Object Detection: ruclips.net/p/PL1GQaVhO4f_jLxOokW7CS5kY_J1t1T17S
If you found this tutorial useful, please share with your friends(WhatsApp/iMessage/Messenger/WeChat/Line/KaTalk/Telegram) and on Social(LinkedIn/Quora/Reddit),
Tag @cogneethi on twitter.com
Let me know your feedback @ cogneethi.com/contact
Amazing tutorials.
Please keep up this great work.
Looking forward for new videos in object detection such as RetinaNet, SSD and more...
Thanks a lot for the encouragement!
Will cover few more papers in a couple of months.
Cheers!
Awesome tutorial. Helps a lot.
Thank You
Firstly. Awesome video series. I am forwarding to lot of my friends. Correct me if I am wrong. When we do Anchor Box processing in the RPN, we also slide 1 pixel at a time (over the CNN feature maps) and evaluate 9 boxes, right ? So, that's also kind of feature pyramid. But here, you are commenting that such feature pyramid is a time consuming idea !!
Avisek,
I think my usage of the term 'Feture Pyramid' here is probably not correct.
Here, what I meant was, if we had used Feature Pyramid, then we had to do ROI pooling using 9 different sliding window sizes at each pixel of Feature Map.
Instead, by using Anchor Boxes, we do ROI pooling using only 1 square sliding window at each pixel of FM and get 9 different predictions. So, that way we will be saving processing time.
Perhaps, it is more clear after watching next video 'C.8.1' and also 'C.8.6 Quirks...'
Let me know if I need to elaborate.
Let me know if something is not clear or if I have made any mistakes too.
Btw, thank you for sharing with your friends!
great work..Keep going!
Hey bro great work make some videos on Mask RCNN and instance segmentation
Thank you. Will do in few months.
your pronounce make me crazy