Thread by @ai_fast_track

🎉🎊🥳Day 30/30: Labeling data and training object detection models are time consuming and expensive.

Here is a Survey of Self-Supervised and Few-Shot Object Detection (FSOD). Those technique aim at alleviating those issues.

The authors have categorized, reviewed, and compared several few-shot and self-supervised object detection methods

📌 FSOD is about training a model on novel (unseen) object classes with little data. It still requires prior training on many labeled data of base (seen) classes

📌 Self-Supervised Learning (SSL) methods aim at learning representations from unlabeled data which transfer well to downstream tasks such as object detection

📌 Combining few-shot and self-supervised object detection is a promising research direction

The authors summarized their main takeaways, made future best practice recommendations, highlighted trends to follow, and given pointers to related tasks.

🔎 TAKEAWAYS & TRENDS

📌 Finetuning is a strong baseline

📌 Impact of self-supervision for object detection

📌 Using heuristics to generate weak labels (e.g. data augmentation)

📌 Rise of transformers

📌 Problems with current evaluation procedures

🔗 RELATED TASKS

📌 Weakly-supervised object detection

📌 Self-supervision using other modalities

📌 Low-data object detection

📌 Few-shot semantic segmentation

📌 Zero-shot object detection

📰 Paper: A Survey of Self-Supervised and Few-Shot Object Detection

abs: arxiv.org/abs/2110.14711
pdf: arxiv.org/pdf/2110.14711…

⭐️ If you find this thread helpful, feel free to follow
@ai_fast_track for more OD / CV demystified content in your feed

⭐️ If you could give this thread a quick retweet, it would help others discover this content. 🙏

https://twitter.com/ai_fast_track/status/1465773297837215750