Global vehicle insurance & vehicle rental industries still rely on manual ways to detect the vehicle damage & its intensity. Visual quality inspection is commonly used for detecting the damage for claim process. The industry is steeped with manual processes, paper-driven operations, high premium offerings, poor customer service, long turnaround time, etc. Here we will use machine learning - object detection “Efficientdet” model with sagemaker and tensor flow. Object detection model will be used to identify & mark the dent and scratch area in the car images.
Let’s refresh the basic terms used in building this ML Model.
Machine learning is a method of data analysis that automates analytical model building. It is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention.
Object detection is a computer technology related to computer vision and image processing that deals with detecting instances of semantic objects of a certain class (such as humans, buildings, or cars) in digital images and videos.
EfficientDet is an object detection machine learning model, which utilizes several optimization and backbone tweaks, such as the use of a BiFPN, and a compound scaling method that uniformly scales the resolution, depth and width for all backbones, feature networks and box/class prediction networks at the same time.
Loss functions is a crucial factor that affecting the detection precision in object detection task. This loss will help with any task which requires classification. We are given k categories and our job is to make sure our model is good job in classifying x number of examples in k categories. Let’s take example of this project where we are given 100 images of 2 categories and our task is to classify each given image into either of these categories “dent” and/or “scratch”.
In this repository, we will build a custom model using Sagemaker & tensorflow to provide bounding boxes on car images consisting of “dents” and/or “Scratch”. Firstly, use Amazon SageMaker Ground Truth to label the car images with bounding box using private workforce option. After finishing the labelling job, ground truth will create & save a manifest file in S3.
Next steps, use Amazon SageMaker to build, train, and deploy an EfficientDet model using the TensorFlow Object Detection API. It is built on top of TensorFlow 2 that makes it easy to construct, train and deploy object detection models. It also provides the TensorFlow 2 Detection Model Zoo which is a collection of pre-trained detection models we can use to accelerate our Model building.
• Label the car images with bounding boxes as “dent” and/or “scratch” using Sagemaker Ground Truth • Generate the dataset TFRecords and label map using SageMaker Processing job • Fine-tune an EfficientDet model with TF2 on Amazon SageMaker • Monitor your model training with Tensorboard and SageMaker Debugger • Deploy your model on a SageMaker endpoint and visualize the prediction by detecting "dent" and/or “scratch” in car images (refer below images)
Follow the step-by-step guide by executing the notebooks in the following folders:
![]() |
![]() |
![]() |
![]() |
This library is licensed under the MIT-0 License. See the LICENSE file.