essamgoda/DeepVO

DeepVO Implementation with PyTorch

DeepVO

• Unofficial PyTorch Implementation of DeepVO: Towards end-to-end visual odometry with deep Recurrent Convolutional Neural Networks
• Pretrained FlowNet weights from ClementPinard/FlowNetPytorch
• RAdam Optimizer from LiyuanLucasLiu/RAdam
• Kitti Dataset
• Midterm for ECE471 Deep Learning Cooper Union

Usage

# Download Kitti Dataset from http://www.cvlibs.net/datasets/kitti/eval_odometry.php
# Download pretrained FlowNet weight
# Edit src/params.py to change hyperparameter and paths 
python3 src/main.py

Results

Training Sequence

• sequence 00
  -
• sequence 02
  -
• sequence 08
  -
• sequence 09
  -

Test Sequence

• sequence 01
  -
• sequence 03
  -
• sequence 05
  -
• sequence 06
  -
• sequence 07
  -
• sequence 10
  -