Goals
The goal was to create a no-code platform that enables novice ML users to quickly prototype and test various ML algorithms for their use cases. This solution also supports deployment onto embedded systems, making it easier to obtain real-world test results.
Solution
SDK is a collection of various ML algorithms for use cases such as classification, segmentation, and object detection. It allows users to specify and configure every part of the training process through configuration files. By specifying the target FPGA, the deployment pipeline handles the necessary transformations to port the ML model to the target FPGA. It utilizes quantization aware training (QAT) to streamline the model. After training, additional transformations are carried out to describe the model into hardware compatible layers which can be deployed onto FPGAs.
Results
- No-code platform that enables quick prototyping
- Easier deployment to supported FPGAs
