3DObjectDetectionPytorch: A 3D Object Detection Model

elegirl_admin

ailia.AI

ailia.AI Editorial Team

Here’s an introduction to “3DObjectDetectionPytorch,” a machine learning model available for use with ailia SDK. By utilizing the ailia SDK and ailia MODELS, which are edge-oriented inference frameworks, implementing AI functionality into applications becomes straightforward.

Overview of 3DObjectDetectionPytorch:

3DObjectDetectionPytorch is a machine learning model designed to compute 3D bounding boxes. While models like YOLO typically compute 2D bounding boxes, 3D ObjectDetectionPytorch calculates 3D bounding boxes that include depth information.

Source: Objectron Dataset

GitHub – sovrasov/3d-object-detection.pytorch

This project provides code to train a two stage 3d object detection models on the Objectron dataset. Training includes…

github.com

Architecture of 3DObjectDetectionPytorch:

3DObjectDetectionPytorch is capable of recognizing nine classes:

OBJECTRON_CLASSES = (‘bike’, ‘book’, ‘bottle’, ‘cereal_box’, ‘camera’, ‘chair’, ‘cup’, ‘laptop’, ‘shoe’)

Initially, MobileNetV2 SSD computes 2D bounding boxes for objects, followed by the computation of 3D bounding boxes using the regression_model of MobileNetV3. The regression_model outputs nine (x, y) keypoints for each class in the format of (9,1,9,2), taking (1,3,224,224) images as input.

3D Object Detection Pytorch is trained on the Objectron dataset provided by Google.

Source:https://github.com/google-research-datasets/Objectron

GitHub — google-research-datasets/Objectron: Objectron is a dataset of short, object-centric video…

Objectron is a dataset of short, object-centric video clips. In addition, the videos also contain AR session metadata…

github.com

The Objectron dataset, developed for augmented reality (AR) purposes, contains 15K annotated video clips and 4M annotated images.

Source:https://github.com/google-research-datasets/Objectron

Using 3DObjectDetectionPytorch:

To use 3DObjectDetectionPytorch, use the following command. It is capable of recognition from a webcam.

$ python3 3d-object-detection.pytorch.py -v 0

Example:

https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FF8jjOikMFpQ&display_name=YouTube&url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DF8jjOikMFpQ&image=http%3A%2F%2Fi.ytimg.com%2Fvi%2FF8jjOikMFpQ%2Fhqdefault.jpg&key=a19fcc184b9711e1b4764040d3dc5c07&type=text%2Fhtml&schema=youtube

ailia-models/object_detection_3d/3d-object-detection.pytorch at master · axinc-ai/ailia-models

(Image from Objectron Dataset…

github.com

Related Information

In ailia SDK, you can also utilize Google’s mediapipe_objectron model, trained on the Objectron dataset.

ailia-models/object_detection_3d/mediapipe_objectron at master · axinc-ai/ailia-models

(Image from Objectron Dataset…

github.com

AX Inc. is a company specializing in practical AI solutions, developing ailia SDK for cross-platform, GPU-accelerated inference. From consulting to model creation, SDK provision, AI-based app/system development, and support, AX Inc. offers a comprehensive solution for AI-related needs. Feel free to contact us for inquiries.