Yolo_v_3

Simplest, easiset implementation of Yolov3 from scratch.

To run the detector on sample directory:

• clone this repository
• save your images in the imgs folder
• download the yolov3 weights using command given below
• run the command python3 detect.py in the terminal
• result will be present in the det folder

To run the detection on video file:

• save your video file in the current directory
• download the yolov3 weights using command given below
• run the command python3 video.py in the terminal

To run the detection on webcam:

• change the code (Cooment the line 94 and Uncomment the line 96) in the video.py
• download the yolov3 weights using command given below
• run the command python3 video.py in the terminal

Command to download Pretrained weights for yolov3:- wget https://pjreddie.com/media/files/yolov3.weights

Results:

Experiment on the thermanl image (flir1.jpeg) from the FLIR thermal image dataset. Detection is present in the det folder (det_flir1.jpeg).

Showing the sample results after the detection (in det folder) on the images (in imgs folder):

Time taken (in seconds) for detection : ``

Reading addresses : 0.000

Loading batch : 0.091

Detection (10 images) : 2.541

Output Processing : 0.000

Drawing Boxes : 0.102

Average time_per_img : 0.273 ``

pallete : For the bounding box of different objects in a image we need different color of bounding boxe.For this we use this pickle file that contains many colors to randomly choose from.

Prediction Visualizations:

Detection on normal image:

Detection on thermal image: