Naimul Mefraz Khan
Embedded Systems
With the rise of social media apps such as Instagram and TikTok, amateur filmmaking has received significant attention. A key aspect in filmmaking is keeping an object of interest in focus. Automating such a process would help in creating spontaneous dynamic videos (e.g. a pet running around or a ball being thrown around the yard).
The objective is to build a dynamic camera that can track an object, keeping it at the center of focus, while livestreaming the captured video through a web application.
1. A camera rig that has 3 degrees of freedom to look around for an object.
2. Once an object is detected, the camera tracks the object and keeps it in focus (center of screen).
3. The camera can move smoothly to keep the object in focus.
4. Live video captured through the camera should be streamed to a web platform.
1. Build a rig with 3dof to mount the camera.
2. Use NVIDIA Jetson Nano or Raspberry Pi to perform on device machine learning that can detect objects (e.g. YoloV3).
3. Track and rotate camera with relation to the movement of the object.
4. Live stream video captured through the camera to a web application.
6. Web application should have option to control the camera (free scanning mode, tracking mode etc).
See Suggested Approach
Build 3dof rig for camera with stepper motors to control movement. Movement should be smooth so that the captured movie isn't "jerky".
Study object detection models, implement it on device
Study object detection models, implement it on device
Streaming video, web application.
NMK07: Object tracking camera rig for dynamic video content creation | Naimul Mefraz Khan | Tuesday August 24th 2021 at 05:16 PM