robar
Published

Radar based 3D Object Detection for mobile robot

Running Center-based Radar and Camera Fusion for 3D Object Detection on KV260

4
Radar based 3D Object Detection for mobile robot

Things used in this project

Hardware components

Kria KV260 Vision AI Starter Kit
AMD-Xilinx Kria KV260 Vision AI Starter Kit
×1
Texas Instruments IWR6843AOPEVM
×1
USB Webcam ZC-D2, Trademark Teaosiy
×1
Digilent DC Motor/Gearbox (1:19 Gear Ratio): Custom 12V Motor
×2
Pmod HB5
Digilent Pmod HB5
×2
Pmod IOXP
Digilent Pmod IOXP
×1
Digilent 2x6-pin to Dual 6-pin Pmod Splitter Cable
×1

Software apps and online services

Texas Instruments Code Composer Studio IDE
Texas Instruments MMWAVE-SDK
Texas Instruments Industrial mmWave toolbox
Texas Instruments mmWave Demo Visualizer
nuScenes
Netron
Eclipse IDE for Embedded C/C++ Developers
PyCharm

Story

Read more

Custom parts and enclosures

Base plate

Source:
https://www.instructables.com/Setting-Up-the-Zybot-Hardware-Round-and-Tall-Editi/

Schematics

dummy

There are no schematics yet.

Code

torch_script_writer.py

Python
output a list of dict in case of multi head outputs of the model because it's currently reduces to the last dict tensor from inference
replace the corresponding function in:
tools/Vitis-AI-Quantizer/vai_q_pytorch/pytorch_binding/pytorch_nndct/export/torch_script_writer.py
or in the conda environment of vitis-ai-pytorch (docker-container)
def _write_forward(self, f: Callable, graph: Graph):
    indent_str = 4 * " "
    f.write('\n' + indent_str + "def forward(self, *args):\n")
    indent_str += indent_str
    self._collect_reuse_output(graph)
    for node in graph.nodes:
      forward_str, output_str = self._get_forward_str(node)
      format_forward_str = self._append_indent(indent_str, forward_str)
      f.write(format_forward_str + '\n')

    return_str = indent_str + 'return [{'
    for i, end_tensor in enumerate(graph.end_tensors):
      if i > 0:
        return_str = ', '.join(
            [return_str, str(i) + ':' + self.get_output_tensor_name(end_tensor)])
      else:
        return_str += str(i) + ':' + self.get_output_tensor_name(end_tensor)
       
    f.write(return_str + '}]\n')

CenterTrackCustom

Cloned from https://github.com/xingyizhou/CenterTrack Contains code changes for Vitis quantization in the separate branch "Vitis-AI_quantization"

iwr6843aop

Driver library to access to iwr6843AOPEVM via UART

CenterTrack

Original repository reference by the paper.

xilinx-k26-starterkit-2021_1

custom petalinux which contains Vitis-AI 2.0 libraries and tools

CenterFusion

Original repository referenced in paper

CenterFusionCustom

Adapt/change source code of the original for Vitis-AI quantization purpose

Credits

robar

robar

1 project • 0 followers

Comments

Add projectSign up / Login