Video Object identification and tracking with customized model
Budget: $750 – $1,500 USD
This project is about object tracking and tagging pilot (after this we will continue with next phases) :
Requirements:
1. connect to live video camera stream of a shed/structure with fixed view (e.g. camera not moving). then you need to:
1. Identify and track the movement of the objects in the shed
2. Count each time an object enter/exit into a tagged area in the picture (by looking at area pixel location and areas will be given in XML)
3. another task is to recognize the image for certain situation.
Instructions:
1. You will get an AWS EC2 access where you run Yolo , tensorflow, deepSORT, torch or other
2. You will use customized model (augmented with public model), the customized model include annotations we labeled in CVAT, you will need to help guide us and customize your self the labels for the model to be accurate and add more with us as needed for getting identification accuracy
3. You will get file with co-ordinates of the tagged areas in the picture , so you can look into it when tracking the object if they enter or exist the areas
4. you will have ~3 types of objects to video track and 2 fixed image to recognize and store the events in database table and google sheet. Need to train, test and run. connect to live camera and show the tracking on video, also save to database when event occur.
5. Have system configuration table that hold all camera settings, areas, and instructions how to run and track an store the data
6. At end to package it: so I can train , test and run in few clicks.
Requirements:
1. connect to live video camera stream of a shed/structure with fixed view (e.g. camera not moving). then you need to:
1. Identify and track the movement of the objects in the shed
2. Count each time an object enter/exit into a tagged area in the picture (by looking at area pixel location and areas will be given in XML)
3. another task is to recognize the image for certain situation.
Instructions:
1. You will get an AWS EC2 access where you run Yolo , tensorflow, deepSORT, torch or other
2. You will use customized model (augmented with public model), the customized model include annotations we labeled in CVAT, you will need to help guide us and customize your self the labels for the model to be accurate and add more with us as needed for getting identification accuracy
3. You will get file with co-ordinates of the tagged areas in the picture , so you can look into it when tracking the object if they enter or exist the areas
4. you will have ~3 types of objects to video track and 2 fixed image to recognize and store the events in database table and google sheet. Need to train, test and run. connect to live camera and show the tracking on video, also save to database when event occur.
5. Have system configuration table that hold all camera settings, areas, and instructions how to run and track an store the data
6. At end to package it: so I can train , test and run in few clicks.