Object Recognition: GitHub Repo + Whitepaper for ‘Finding Tiny Faces’

We describe a detector that can find ~800 faces out of the reportedly 1000 present, by making use of novel characterization of scale, resolution, and context to find small objects. Detector confidence is given by the color bar on the right. Can you confidently identify errors?

Though tremendous strides have been made in object recognition, one of the remaining open challenges is detecting small objects. We explore three aspects of the problem in the context of finding small faces: the role of scale invariance, image resolution, and contextual reasoning. While most recognition approaches aim to be scale-invariant, the cues for recognizing a 3px tall face are fundamentally different than those for recognizing a 300px tall face. We take a different approach and train separate detectors for different scales. To maintain efficiency, detectors are trained in a multi-task fashion: they make use of features extracted from multiple layers of single (deep) feature hierarchy. While training detectors for large objects is straightforward, the crucial challenge remains training detectors for small objects. We show that context is crucial, and define templates that make use of massively-large receptive fields (where 99% of the template extends beyond the object of interest). Finally, we explore the role of scale in pre-trained deep networks, providing ways to extrapolate networks tuned for limited scales to rather extreme ranges. We demonstrate state-of-the-art results on massively-benchmarked face datasets (FDDB and WIDER FACE). In particular, when compared to prior art on WIDER FACE, our results reduce error by a factor of 2 (our models produce an AP of 81% while prior art ranges from 29-64%).

See the GitHub repo here, read more here and read whitepaper here.


Have an amazing project to share? The Electronics Show and Tell is every Wednesday at 7:30pm ET! To join, head over to YouTube and check out the show’s live chat and our Discord!

Join us every Wednesday night at 8pm ET for Ask an Engineer!

Join over 38,000+ makers on Adafruit’s Discord channels and be part of the community! http://adafru.it/discord

CircuitPython – The easiest way to program microcontrollers – CircuitPython.org


New Products – Adafruit Industries – Makers, hackers, artists, designers and engineers! — New Products 11/15/2024 Featuring Adafruit bq25185 USB / DC / Solar Charger with 3.3V Buck Board! (Video)

Python for Microcontrollers – Adafruit Daily — Python on Microcontrollers Newsletter: A New Arduino MicroPython Package Manager, How-Tos and Much More! #CircuitPython #Python #micropython @ThePSF @Raspberry_Pi

EYE on NPI – Adafruit Daily — EYE on NPI Maxim’s Himalaya uSLIC Step-Down Power Module #EyeOnNPI @maximintegrated @digikey

Adafruit IoT Monthly — The 2024 Recap Issue!

Maker Business – Adafruit Daily — Apple to build another chip at TSMC Arizona

Electronics – Adafruit Daily — SMT Tip – Stop moving around!

Get the only spam-free daily newsletter about wearables, running a "maker business", electronic tips and more! Subscribe at AdafruitDaily.com !


No Comments

No comments yet.

Sorry, the comment form is closed at this time.