{"id":2486,"date":"2021-09-21T09:00:31","date_gmt":"2021-09-21T01:00:31","guid":{"rendered":"https:\/\/blog.ailia.ai\/?p=2486"},"modified":"2025-05-20T16:26:34","modified_gmt":"2025-05-20T08:26:34","slug":"3dobjectdetectionpytorch-3d-object-detection-model","status":"publish","type":"post","link":"https:\/\/blog.ailia.ai\/en\/tips-en\/3dobjectdetectionpytorch-3d-object-detection-model\/","title":{"rendered":"3DObjectDetectionPytorch : 3D Object Detection Model"},"content":{"rendered":"\n

Overview<\/strong><\/h3>\n\n\n\n

3DObjectDetectionPytorch<\/em>\u00a0is a machine learning model that calculates 3D bounding boxes of objects. Other object detection models such as\u00a0YOLO<\/a>generally computes 2D bounding boxes, but this new model returns bounding boxes with depth information.<\/p>\n\n\n\n

\"\"
Source: Objectron dataset<\/figcaption><\/figure>\n\n\n\n

<\/a>GitHub – sovrasov\/3d-object-detection.pytorch<\/a><\/p>\n\n\n\n

Architecture<\/strong><\/h3>\n\n\n\n

3DObjectDetectionPytorch<\/em>\u00a0is capable of recognizing the following 9 classes.<\/p>\n\n\n\n

\n

OBJECTRON_CLASSES = (\u2018bike\u2019, \u2018book\u2019, \u2018bottle\u2019, \u2018cereal_box\u2019, \u2018camera\u2019, \u2018chair\u2019, \u2018cup\u2019, \u2018laptop\u2019, \u2018shoe\u2019)<\/p>\n<\/blockquote>\n\n\n\n

First, the 2D bounding box of the object is computed with MobileNetV2 SSD<\/em>, and then the 3D bounding box is calculated with MobileNetV3<\/em> regression model. This regression model takes an image (1,3,224,224) as input and returns 9 keypoints (x,y) for each class (9,1,9,2).<\/p>\n\n\n\n

3DObjectDetectionPytorch<\/em> was trained on the Objectron<\/em> dataset, which is publicly available from Google<\/em>.<\/p>\n\n\n\n

\"\"
Source: https:\/\/github.com\/google-research-datasets\/Objectron<\/a><\/figcaption><\/figure>\n\n\n\n

<\/a>GitHub \u2014 google-research-datasets\/Objectron<\/a><\/p>\n\n\n\n

The\u00a0Objectron\u00a0<\/em>dataset is a dataset developed for AR development and contains 15K annotated videos and 4M annotated images.<\/p>\n\n\n\n

\"\"
Source: https:\/\/github.com\/google-research-datasets\/Objectron<\/a><\/figcaption><\/figure>\n\n\n\n

Usage<\/strong><\/h3>\n\n\n\n

Use the following command to run\u00a03DObjectDetectionPytorch\u00a0<\/em>with ailia SDK on a web camera video stream.<\/p>\n\n\n\n

$ python3 3d-object-detection.pytorch.py -v 0<\/code><\/p>\n\n\n\n

Here is the result you can expect.<\/p>\n\n\n\n

\n