Summary of A Partial Replication Of Maskformer in Tensorflow on Tpus For the Tensorflow Model Garden, by Vishal Purohit et al.
A Partial Replication of MaskFormer in TensorFlow on TPUs for the TensorFlow Model Garden
by Vishal Purohit, Wenxin Jiang, Akshath R. Ravikiran, James C. Davis
First submitted to arxiv on: 29 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG); Software Engineering (cs.SE)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper replicates the MaskFormer model, a universal image segmentation model originally developed using PyTorch and optimized for execution on TPUs within TensorFlow. The implementation leverages the modular constructs in the TensorFlow Model Garden (TFMG), including data loaders, training orchestrators, and architectural components, tailored to meet MaskFormer’s specifications. Key challenges include non-convergence issues, slow training, adapting loss functions, and integrating TPU-specific functionalities. Qualitative results are presented on the COCO dataset, verifying the reproduced implementation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper copies a special kind of computer model called MaskFormer, which helps computers understand what’s in pictures. The team working on this project made the original model work better for special computers called TPUs. They used some special tools and steps to make it happen. Some problems they faced included slow learning, not getting the right answers, and figuring out how to use these special computers. In the end, their new version works well enough to show what it can do on a famous dataset. The code for this project is available online. |
Keywords
» Artificial intelligence » Image segmentation
