Imitation learning based on Bird-eye view for follow lane autonomous driving in CARLA simulator
We explore learning a policy for end-to-end driving in CARLA
Logbook
2022
Carla simulator maps
List of available maps
Time cycle experiments analysis
Analysis on 3 circuits
Time cycle experiments analysis
Analysis on 3 circuits
Training models with weather changes
Comparing performance with new training procedure
Analyzing Neural paper experiment with new dataset
First experimental results analysis
Analyzing Neural paper experiment with old dataset
First experimental results analysis
Analyzing experiments classification vs regression
Different classification and regressions brains used.
Wrapping up ideas
Wrapping up ideas
Opencv new dataset vs Explicit dataset vs old dataset comparison
Comparing how the brains perform training with different datasets
New dataset vs old dataset comparison
Comparing how the brains perform training with different datasets
Looking at the distribution of cases in the dataset
Plotting the cases distribution of the dataset
Extending new dataset and training Pilotnet model
The new dataset has been expanded for a broader support.
Comparing cropping strategies
Experiments on cropping
Experiments analysis for simple circuit and Montmeló
Looking at some experiments
Paper first approach experiments analysis
Looking at some experiments
Review of current neural architectures used
Reviewing current neural architectures to set the base fot following steps.
Comparing retrained Pilonet vs Pilotnet3D on extreme difficulty circuits
Comparing advanced brains on difficult circuits
Brain iterations comparison with updated metrics (last experiment rerun)
Comparing how changing the brain iterations affects the output with updated metrics (last experiment rerun)
Real time factor comparison with new metrics
Comparing how changing the real time factor plays a role in the performance
Brain max speed comparison
Comparing how changing the max brain speed affects the output
Brain iterations comparison with updated metrics
Comparing how changing the brain iterations affects the output with updated metrics
Brain iterations comparison extended
Comparing how changing the brain iterations affects the output
Montmeló vs simple circuit no red line comparison fot all the brains
Comparing all the different brains developed on two circuits
Brain iterations comparison
Comparing how changing the brain iterations affects the output
Max speed comparison
Comparing how changing the speed of the robot affects the output
Real time factor comparison
Comparing how changing Gazebo simulation real time factor affects the results
2021
Where are the extreme cases located on the circuits?
I look for the extreme cases considering the angular speed and derivative value on the PID controller.
New PID brain added to the comparison
New PID brain added to the comparison
New position deviation calculation
New position deviation calculation
Experiments with DeepestLSTMTinyPilonet and Pilotnet on more circuits
Experiments with DeepestLSTMTinyPilonet and Pilotneton more circuits
Experiments with DeepestLSTMTinyPilonet on more circuits
Experiments with DeepestLSTMTinyPilonet on more circuits
Comparing Pilotnet vs Deepest LSTM TinyPilotNet using Behavior Metrics
Comparison between two brains that complete all the circuits in Behavior Metrics.
Applying data augmentations with Albumentations on Tensorflow
Applying data augmentations with Albumentations on Tensorflow
Preprocessed images with and without memory (LSTMs)
Preprocessed images experiments using basic NNs and LSTMs
Exploring data augmentation for LSTMs
Exploring data augmentation for LSTMs
More experiments adding classification and extreme sequence data for LSTM models
More experiments with a new classification model and extreme data for the LSTM based models
Improving previous experiments
Taking last week’s experiments and improving the results
Learning to drive with deep learning. Experiments with 4 types of nets and comparison
Current state of the project and experiments made
2020
Behavior Studio current state [oct 2020]
Current state of the project and experiments made during last weeks.
Web update, Behavior Studio installation
This webpage is updated with blog posts from previous weeks and general LSTM paper reading is conducted.
Web update, Behavior Studio installation
This webpage is updated with blog posts from previous weeks and general LSTM paper reading is conducted.
Paper writing, documentation update and RetinaNet reading
The paper is completed and RetinaNet paper is read.
Object detection models review
General review of object detection papers to improve the paper and the experiments.
Network models testing
Experiments with some of the most common object detection models are conducted.
PyTorch inferencer included
PyTorch framework support is included in Detection Studio.
OpenCV-Yolov3 inferencer included
OpenCV-Yolov3 support is included in Detection Studio, getting rid of darknet framework dependency.