As discussed earlier (here and here), and based on the way the segmentation is being used in the WeedZSL repository, we decided to use this approach as the preprocessing step before the nvinfer inference.
There're another approaches of doing that except for using Excess Green Index. Please try to do the research and propose alternative approaches.
As the outcome of the ticket, please prepare the documentation where you provide code examples, and compare different approaches (based on libs requirements, complexity, possibility to port to GPU, visual quality with examples), and measure the CPU latency (e.g., on one of the test images)
DoD:
- Detailed readme (with examples) created
As discussed earlier (here and here), and based on the way the segmentation is being used in the WeedZSL repository, we decided to use this approach as the preprocessing step before the nvinfer inference.
There're another approaches of doing that except for using Excess Green Index. Please try to do the research and propose alternative approaches.
As the outcome of the ticket, please prepare the documentation where you provide code examples, and compare different approaches (based on libs requirements, complexity, possibility to port to GPU, visual quality with examples), and measure the CPU latency (e.g., on one of the test images)
DoD: