Inria STARS Deep Learning for Computer Vision
Winter 2023
[Home | Schedule | Final Project ]

Project Topics

We suggest 1 or 2 people as a team

Option 1: Choose from the suggested papers (updating)

• Learning Deep Features for Discriminative Localization (CVPR 2016)
• Beyond Part Models: Person Retrieval with Refined Part Pooling (and A Strong Convolutional Baseline) (ECCV 2018)
• Repulsion Loss: Detecting Pedestrians in a Crowd (CVPR 2018)
• An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Data (AAAI 2017)
• SlowFast Networks for Video Recognition (ICCV 2019)
• Long Short-Term Transformer for Online Action Detection ( NeurIPS2021)
• Real-time Online Video Detection with Temporal Smoothing Transformers (ECCV 2022)
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (ICCV 2017)
• Image-to-Image Translation with Conditional Adversarial Networks (CVPR 2017)
• Emerging Properties in Self-Supervised Vision Transformers (ICCV 2021)

Option 2: Your own project.

You can also pick papers from CVPR/ICCV/ECCV/NeuRIPS/ICLR/ICML. You also let us know before...

Grading Policy

Final project: 100%

Tips on how to read a machine learning scientific paper

(as stated in the slides from the first lecture):

• Stanford CS230: Career Advice / Reading Research Papers (from 0:59 to 28:51)
• Summarized in text

Tips on writing a paper review

(Based on example review forms, from the 2020 AAAI conference) Address the following parts:

• Motivation/Relevance: why did the authors do what they did?
• Related work/Novelty: how is this situated in the research field/current state-of-the-art?
• Contribution: what is the presented contribution exactly, what does it do and how does it work?
• Support/Evaluation/Correctness: how are the claims supported: theory and/or empirical evaluation?
• Impact/Significance: how will this work change the world?
• Writing/Clarity: how easy was it to read?

For most of these categories, it is good to both highlight the authors' view and your own interpretation on it. Exemplary format:

• 1. [Summary] Please summarize the main claims/contributions of the paper in your own words.
• 2. [Relevance] Is this paper relevant to an AI audience?
• 3. [Significance] Are the results significant?
• 4. [Novelty] Are the problems or approaches novel?
• 5. [Soundness] Is the paper technically sound?
• 6. [Evaluation] Are claims well-supported by theoretical analysis or experimental results?
• 7. [Clarity] Is the paper well-organized and clearly written?
• 8. [Detailed Comments] Please elaborate on your assessments and provide constructive feedback.
• 9. [QUESTIONS FOR THE AUTHORS] Please provide questions for authors to address during the author feedback period. (E.g. if something is unclear to you)