Robotic studio 2- Autonomous speed traffic sign detection and implementation

Week-1 (18/05/2020 – 24/05/2020)

Monday(18/05/2020)

1. Preparing for project sprint that is to be held on 19/05/2020
2. I have to complete and demonstrate the 50% milestone of my project. This includes the duckiebot being able to autonomously drive within lanes of the duckietown and detects traffic signs.
3. I had earlier used YOLO detection, using custom weights. It is not working properly so, I am going to try doing it with CNN whose code I found online. For the presentation, both of my techniques are working fine. (Yolo package used- https://github.com/lasithaya/Robotics-Studio-2/wiki/Week6-:-Train-the-YOLO-object-detector)

Wednesday(20/05/2020)

The project sprint went well. Now the next step is to complete the next mile stone of the project and start with peer teaching presentation.

Thursday(21/05/2020)

There was a project help class, a 15 min session for asking questions and discussing the project. Preparing for Peer teaching that is to be held on 26/05/2020. The topics I choose to cover are:

• TSR- Traffic Sign Recognition
• Traffic Sign Detection using CNN
• YOLO object detector

Sturday(23/05/2020)

1. TSR- Traffic Sign Recognition

TSR is a visual-based traffic sign recognition system that can be implemented on the automobile with an aim of detecting and recognizing all emerging traffic signs. The same would be displayed to the driver with alarm-triggering features if the driver refuses to follow the traffic signs. This system includes following steps,

• Detection: Blurring, Colour-based detection, Shape-based detection, Cropped and extract features, Separation

• Recognition: Tesseract, SVM, Tempelate matching, ORB

2. Traffic Sign Detection using CNN

A Convolutional Neural Network (ConvNets/CNN) are very similar to ordinary neural networks. They are made up of neurons with learnable weights and biases. All the tips and tricks learned or used for ordinary neural networks are also applicable to CNN’s.

The only difference between CNN and the ordinary neural network is that CNN assumes that input is an image. This vastly reduces the number of parameters to be tuned in a network.

• Convolutional layer
• Pooling Layer
• Fully-connected Layer (FC layer)

3. Yolo object detector

YOLO is a clever convolutional neural network (CNN) for doing object detection in real-time. The algorithm applies a single neural network to the full image, and then divides the image into regions and predicts bounding boxes and probabilities for each region. These bounding boxes are weighted by the predicted probabilities. YOLO divides the input image into an S×S grid. Each grid cell predicts only one object.

If the center of an object falls into a grid cell, that grid cell is responsible for detecting that object. Each grid cell predicts B bounding boxes and confidence scores for those boxes. If no object exists in that cell, the confidence scores should be zero. Each bounding box consists of 5 predictions: x, y, w, h, and confidence. The width and height are predicted relative to the whole image. Each grid cell also predicts C conditional class probabilities. These probabilities are conditioned on the grid cell containing an object. Only one set of class probabilities is predicted per grid cell, regardless of the number of boxes B.

Sunday(24/05/2020)

Made the presentation. Decided on the questions I am going to ask afterwards. Will be going back to completing the project after this.