TDA4VM: Object Classification/Tracking and Situation Recognition Algorithm

Hi,

Jack said:

1. Could you please let us know if above features are supported by TI' AI examples?

We support wide range of classic deep learning tasks they are Object Detection, Object Classification, Object Segementation.

I would recommend to read out the required documents here : 

If you are using edgeai linux sdk : https://software-dl.ti.com/jacinto7/esd/processor-sdk-linux-sk-tda4vm/09_01_00/exports/edgeai-docs/common/sdk_overview.html

If you using OSRT : https://github.com/TexasInstruments/edgeai-tidl-tools

Jack said:

2. Which AI model should they refer to to to develop algorithms such as dangerous situation detection and traffic accident detection?

This is dependent on what kind of use case they are working with, we support std models like Mobilenet, yolo, resnet 50 and much more you can refer here : https://github.com/TexasInstruments/edgeai-modelzoo

However please note that, customer needs to do due diligence for model training and model selection as part of TI offering we support model inference on TI SoCs.

Jack said:

3. Please recommend a product that supports performance and examples suitable for processing the above algorithms.

Depending on complexity and computing power, you can look at AM68A, AM68PA, AM69A options available from jacinto device family.

Jack said:

4. Differences between TDA4VM and AM69A (feature, performance focus)

I would recommend to file separate e2e for this, so you can get detailed support.

Jack said:

5. Fisheye distortion correction is possible in the ISP or VPAC engine of TDA4VM and AM69A ?

Please file separate e2e, to get more detailed help on this