Download presentation
Presentation is loading. Please wait.
1
Object Detection with Bootstrapping
Week 6 Carlos Rubiano Mentor: Oliver Nina
2
Assisted Meera, Si, and Arjun
Helped with installing Caffe and its dependencies Debugged errors to setup Caffe instructed how to train on their own custom datasets
3
Ensemble of Networks Trained more networks to average together
Resulted in increase of accuracy but plateau is quicker than expected 7th/155 on Kaggle
4
Training with filter on Image
Coded with python way to manipulate the pickled image data Allows us to train with custom filters on cuda - Sobel - Grayscale Frog Automobile
5
Other filters Class: Frog Data augmentation
Also want to try image rotations, image resizing, etc
6
Things to do Implement network in network, state of art (ICLR 2014)
Implement maxout networks Combine with dropconnect Explore other custom features
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.