Hi, I’m Constantin. I recently studied the fastai book, which I enjoyed very much and made me interested in going deeper in DL and CV.
I’ve been searching for a (chronologically-ordered) short-list of cornerstone papers on the topic of “semantic segmentation” (and similar for “image classification”, etc). For example, I’d like to see the most important 10, then 30, then 100 in chronological order, that shaped the field in some meaningful way. A measure for “most important” as number of citations in other papers (or similar) should be fine for this purpose.
Even more specifically, I’d like to read on the evolution of segmentation architectures (UNet, UNet+, DeepLabV1/2/3/3+) and the associated encoders (VGG, then ResNet, then ResNext, then so on and so forth) in their context at the time of publishing.
I wasn’t able to find anything like this. arxiv and google-scholar offer some search capability for it seems quite limited, and there’s still significant filtering and classification remaining to be done manually. My hope is that I’ve been missing a tool/feature somewhere that’ll just answer what I’m looking for.
I appreciate any suggestions on the same, thank you.
Five months later, and I now know there was no such tool out there.
So I’ve built it
If you’re interested, please check it out at https://www.papelist.app
By default you’ll see the most influential papers (by citations) in Computer Vision overall, and you can filter down by title (e.g., “segmentation”), author, year-range, venue.
Soon, support for other domains (beyond Computer Vision) will be added.
Please check it out and drop a line with your feedback (what’s good, what could be better) at admin@papelist.app
I’ve added several additional fields of study since (now 20 total), including “Computation and Language”, “Machine Learning”, “Robotics” and other CS domains most popular on Arxiv.
I very much appreciate any feedback, as I continue to work on this.
Thanks again,
~Constantin
Consider utilizing platforms like Papers with Code, which provide organized collections of papers, code implementations, and benchmarks. They offer search and filtering options based on citations and relevance, aiding your exploration efficiently.
@Pointing7 - Thanks, I had seen it before asking my original question here.
I agree with you that it’s super useful and I also like it too - however, I’m not finding the answer to my specific question there.
Since I asked the question almost a year ago, I’ve build www.papelist.app and solved it for myself and for others. I think I’ve mentioned this in a previous response.
@Pointing7 - when you say “consider utilizing platforms like …”, it seems you had some other platforms in mind that might answer my original question. I’d love to hear about them, feel free to include.
Thank you for sharing.
