How to Train an Accurate and Efficient Object Detection Model on Any Dataset

TL;DR

This paper introduces a dataset-agnostic object detection training pipeline that provides strong, out-of-the-box models suitable for various datasets, reducing the need for extensive fine-tuning and optimizing for performance and efficiency.

Contribution

The authors propose a universal training template with pre-trained models and a robust pipeline, optimized through parallel training on multiple datasets, enabling effective deployment across diverse use cases.

Findings

Identified three top-performing architectures: VFNet, ATSS, and SSD.

Achieved strong baseline results across multiple datasets.

Models can be deployed on CPU using OpenVINO toolkit.

Abstract

The rapidly evolving industry demands high accuracy of the models without the need for time-consuming and computationally expensive experiments required for fine-tuning. Moreover, a model and training pipeline, which was once carefully optimized for a specific dataset, rarely generalizes well to training on a different dataset. This makes it unrealistic to have carefully fine-tuned models for each use case. To solve this, we propose an alternative approach that also forms a backbone of Intel Geti platform: a dataset-agnostic template for object detection trainings, consisting of carefully chosen and pre-trained models together with a robust training pipeline for further training. Our solution works out-of-the-box and provides a strong baseline on a wide range of datasets. It can be used on its own or as a starting point for further fine-tuning for specific use cases when needed. We obtained dataset-agnostic templates by performing parallel training on a corpus of datasets and optimizing the choice of architectures and training tricks with respect to the average results on the whole corpora. We examined a number of architectures, taking into account the performance-accuracy trade-off. Consequently, we propose 3 finalists, VFNet, ATSS, and SSD, that can be deployed on CPU using the OpenVINO toolkit. The source code is available as a part of the OpenVINO Training Extensions (https://github.com/openvinotoolkit/training_extensions}