Skip to content

Commit

Permalink
Browse files Browse the repository at this point in the history
  • Loading branch information
ndrplz committed Oct 20, 2017
2 parents 89e8d28 + bcc2d13 commit 881dd4a
Showing 1 changed file with 34 additions and 6 deletions.
40 changes: 34 additions & 6 deletions capstone_traffic_light_classifier/README.md
Original file line number Diff line number Diff line change
@@ -1,15 +1,43 @@
# Traffic Light Classifier

Very simple traffic light classifier to be integrated in the capstone project.
Simple traffic light classifier to be integrated in the capstone project.

---

## Project Structure
## Dataset

### Simulator dataset
**Number of frames:** 1324 frames

**Resolution:** 800x600 pixel

**Classes:** {none, red, yellow, green}

**Samples:**
![simulator_dataset](img/simulator_examples.png)


### Udacity circuit dataset
**Number of frames:** 1151 frames

The traffic light dataset is modeled in [`traffic_light_dataset.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/traffic_light_dataset.py)
**Resolution:** 800x600 pixel

The deep network architecture is in [`traffic_light_classifier.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/traffic_light_classifier.py)
**Classes:** {none, red, green}

Training the model: [`train.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/train.py)
**Samples:**
![realworld_dataset](img/realworld_examples.png)

### Mixed dataset

The network has been from the dataset obtained merging the two dataset above.

### Downloading the data

All three datasets are available in this [shared folder](https://drive.google.com/open?id=0Bx9YaGcDPu3XSlMxeHlPZEIyRzQ).

## Project Structure

Testing the model from pretrained weights: [`test.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/test.py)
- [`train.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/train.py): entry point for training the model
- [`test.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/test.py): entry point for testing the model from pretrained weights
- [`traffic_light_dataset.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/traffic_light_dataset.py): contains`TrafficLightDataset` class to abstract over the raw data
- [`traffic_light_classifier.py`](https://github.com/ndrplz/self-driving-car/blob/master/capstone_traffic_light_classifier/traffic_light_classifier.py): contains `TrafficLightClassifier` deep network for classification

0 comments on commit 881dd4a

Please sign in to comment.