This is a repo for documents I have done concerning reproductions and analyses on SoTA papers at the moment of their creation in time. Because of differences in Markdown Preview and actual GitHub Visual Interface, there will be a multitude of small corrective commits done on the fly.
The format and extension of the documents are ever-changing and only representative of my personal formatting choices as I try to learn the craft and do a better job. These analyses will range between mere synopses (shorter and more concise) and bold reproduction papers (detailed and dissected).
An observation can be made to the fact that I like to refer to myself in plural, in hopes of finding an interested party in partaking this past-time activity.
Papers reviewed so far:
1. R-CNN, Fast R-CNN and Faster R-CNN (Object detection)
2. YOLO (Real-Time Object detection)
3. YOLACT (Real-Time Instance Segmentation)
4. FCOS (One-Stage Object Detection)
5. CenterMask (Real-Time Anchor-Free Instance Segmentation)
Honorable mentions:
- Pix2PixHD - Image Reconstruction using GANs - reproduction paper used to apply for EEML2019 and reviewed positively, which I have lost in space and time and cannot find on any email, cloud or local drive.
- Pix2Vox - 3D object Reconstruction from RGB single-shot and multi-shot images - the paper that started my pursue of overcoming the SotA on the 3D-R2N2 Dataset
Disclaimer: The documents might include up to 30% of text from the respective paper, as the explanations taken could not be reproduced in better words, and the documents are relevant to the fact that these writings are of personal academic interest. Use of these outside personal interest or otherwise give me zero liability for any event. Moreover, they are written to the best of my academic understanding meaning that misjudgements may appear.