Jonghoon Jin (jhjin0 at gmail)

Deep learning for computer vision at Lighthouse AI


Deep networks robust to noise

Despite of the success of deep networks, they could be easily fooled by few pixels of noise so as to output incorrect answers. Our feedforward model utilizes uncertainty information and achieves high robustness against strong noise on a large-scale dataset.

Video ICLR16 Poster Lua/Torch7

Lightweight deep networks

Since deep networks contain massive and redundant parameters, finding a compact representation is a key to save runtime and memory. Our model based on filter separation can substitute for 3D filters without accuracy loss and provides 2x speed-up with 90% parameter reduction.

ICLR15 Poster Lua/Torch7

Deep learning in your pocket

Deep networks are state-of-the-art models for visual understanding but require off-site server for processing due to heavy computational cost. Our accelerator on a mobile coprocessor is capable of a peak performance of 240 G-ops/s while consuming less than 4 W.

Video Patent TNNLS17 NIPSW13 CVPRW14 MWSCAS14 HPEC14 MIT Tech Review USA Today BBC Business Insider WLFI18 Journal and Courier EE Times Purdue IMPACT