I’ve got the following idea in my mind. I want to detect objects with YoloV8 and OpenCV. Here a simple example:
However, my idea is slightly different. I intend to capture the camera stream, transmit it to AWS Kinesis, and subsequently, within a separate process, ingest this Kinesis stream for object detection or any other purpose using the camera frames. To achieve this, our initial step is to publish our OpenCV frames into Kinesis. producer.py
Now our consumer.py
To get the frames from kinesis stream we need to consider the shards. According to your needs you should change it a little bit.
And that’s all. Real-time Object Detection with Kinesis stream. However, computer vision examples are often simple and quite straightforward. On the other hand, real-world computer vision projects tend to be more intricate, involving considerations such as cameras, lighting conditions, performance optimization, and accuracy. But remember, this is just an example.
Full code in my github.