AI technical projects
Original project reports and technical notes on reinforcement learning, computer vision, competitions, and embodied AI.
Detecting Ancient Ink in the Vesuvius Challenge
Our gold-medal, 8th-place solution to Kaggle’s 2023 Vesuvius ink detection challenge: 3D convolutional networks, SegFormer, depth augmentation, and center-only patch stitching.
Train a Reinforcement Learning Agent to play Football
Kaggle launched a competition that challenged participants to train an AI agent to play football in the Google Research Football Environment. As a big football fan since childhood, I thought this would be a great opportunity for me to apply reinforcement learning (my favorite ML subject!) to a very interesting scenario that I am
Techniques to Tackle Overfitting and Achieve Robustness for Donkey Car Neural Network Self-Driving Agent
Application of Style Transfer as an data augmentation technique to improve model's robustness.
Train Donkey Car in Unity Simulator with Reinforcement Learning
Donkey Car trained with Double Deep Q Learning (DDQN) in Unity Simulator.
Train a Reinforcement Learning agent to play custom levels of Sonic the Hedgehog with Transfer Learning
OpenAI hosted a contest challenging participants to create the best agent for playing custom levels of the classic game Sonic the Hedgehog, without having access to those levels during development. The goal of this competition is to come up with a meta-learning algorithm that can transfer learning from a set of training levels t
Direct Future Prediction - Supervised Learning for Reinforcement Learning
Instruct DFP agent to change objective (at test time) from pick up Health Packs (Left) to pick up Poision Jars (Right). The ability to pursue complex goals at test time is one of the major benefits of DFP.
Distributional Bellman and the C51 Algorithm
I got the chance to read this paper on Distributional Bellman published by DeepMind in July. Glossing over it the first time, my impression was that it would be an important paper, since the theory was sound and the experimental results were promising. However, it did not generate as much noise in the reinforcement learning comm
Deep Q Network vs Policy Gradients - An Experiment on VizDoom with Keras
After a brief stint with several interesting computer vision projects, include this and this, I’ve recently decided to take a break from computer vision and explore reinforcement learning, another exciting field. Similar to computer vision, the field of reinforcement learning has experienced several important breakthroughs made
Detect and Classify Species of Fish from Fishing Vessels with Modern Object Detectors and Deep Convolutional Networks
I recently took part in the Nature Conservancy Fisheries Monitoring Competition organized by Kaggle. Of all the image related competitions I took part before, this is by far the toughest but most interesting competition in many regards. I managed to finish in the 3rd place in public leaderboard but dropped to 8th place in the pr
Convert Caffe weights to Keras for ResNet-152
Those who have applied deep learning would know, being deep is both a curse and blessing. On the one hand, deep models are endowed with higher representation power; on the other hand, the increase in parameters means more data are necessary to train up the models so that they would generalize well to unseen data, to avoid a phen
Lessons learned from Kaggle StateFarm Challenge
I recently took part in the Kaggle State Farm Distracted Driver Competition. This is an image recognition problem which deep learning is particular good at solving. Hence by taking part in this competition I got the perfect opportunity to acquaint myself with the deep learning modeling pipeline, which typically entails data prep
A Comprehensive guide to Fine-tuning Deep Learning Models in Keras (Part II)
This is Part II of a 2 part series that cover fine-tuning deep learning models in Keras. Part I states the motivation and rationale behind fine-tuning and gives a brief introduction on the common practices and techniques. This post will give a detailed step-by-step guide on how to go about implementing fine-tuning on popular mod
A Comprehensive guide to Fine-tuning Deep Learning Models in Keras (Part I)
In this post, I am going to give a comprehensive overview on the practice of fine-tuning, which is a common practice in Deep Learning.
