GOIS
Enhancing Tiny Object Detection Using Guided Object Inference Slicing (GOIS): An Efficient Dynamic Adaptive Framework for Fine-Tuned and Non-Fine-Tuned Deep Learning Models
Install / Use
/learn @MMUZAMMUL/GOISREADME
MIT License -All rights reserved to the Dr.Muzammul. This project may be used for study and educational purposes, but **redistribution, redevelopment, or use of the code for personal or commercial purposes is strictly prohibited without the author's written consent.
🎥 Watch Live Demo (YouTube) | 🎥 Watch Live Demo (Bilibili)
🚀 Enhancing Tiny Object Detection Using Guided Object Inference Slicing (GOIS): An Efficient Dynamic Adaptive Framework for Fine-Tuned and Non-Fine-Tuned Deep Learning Models
Guided-Object Inference Slicing (GOIS) Innovatory Framework with Several Open source code Deployed on Google Colab/Gradio Live/Huggingface- NEW Legendary INNOVATION
🔬 Research by: Muhammad Muzammul, Xuewei Li, Xi Li
📄 Published in Neurocomputing Journal-> https://doi.org/10.1016/j.neucom.2025.130327
Contact: muzamal@zju.edu.cn
📌 Citation
@article{MUZAMMUL2025130327,
title = {Enhancing Tiny Object Detection Using Guided Object Inference Slicing (GOIS): An efficient dynamic adaptive framework for fine-tuned and non-fine-tuned deep learning models},
journal = {Neurocomputing},
volume = {640},
pages = {130327},
year = {2025},
issn = {0925-2312},
doi = {https://doi.org/10.1016/j.neucom.2025.130327},
url = {https://www.sciencedirect.com/science/article/pii/S0925231225009993},
author = {Muhammad Muzammul and Xuewei Li and Xi Li},
keywords = {Tiny Object Detection, Guided Object Inference Slicing (GOIS), Adaptive slicing-based detection, UAV-based real-time inference, High-resolution remote sensing imagery, Computationally efficient object detection, Deep learning for small-object recognition, Non-maximum suppression optimization, Transformer-based object detection},
abstract = {Tiny Object Detection (TOD) in UAV and standard imaging remains challenging due to extreme scale variations, occlusion, and cluttered backgrounds. This paper presents the Dynamic Adaptive Guided Object Inference Slicing (GOIS) framework, a two-stage adaptive slicing strategy that dynamically reallocates computational resources to Regions of Interest (ROIs), enhancing detection precision and recall. Unlike static and semi-adaptive slicing methods like SAHI and ASAHI, evaluated with models such as FNet, TOOD, and TPH-YOLO, GOIS leverages VisDrone and xView datasets to optimize hierarchical slicing and dynamic Non-Maximum Suppression (NMS), improving tiny object detection while reducing boundary artifacts and false positives. Comprehensive experiments using MS COCO-pretrained Ultralytics models under fine-tuning and non-fine-tuning conditions validate its effectiveness. Evaluations across YOLO11, RT-DETR-L, YOLOv8s-WorldV2, YOLOv10, YOLOv8, and YOLOv5 demonstrate that GOIS consistently outperforms Full-Image Inference (FI-Det), achieving up to 3-4× improvements in small-object recall. On the VisDrone2019 dataset, GOIS-Det improved mAP@0.50:0.95 from 0.12 (FI-Det) to 0.33 (+175%) on YOLO11 and from 0.18 to 0.38 (+111.10%) on YOLOv5n. Fine-tuning further enhanced AP-Small by 278.66% and AR-Small by 279.22%, confirming GOIS’s adaptability across diverse deployment scenarios. Additionally, GOIS reduced false positives by 40%–60%, improving real-world detection reliability. Ablation studies validate GOIS’s hierarchical slicing and parameter optimization, with 640-pixel coarse slices and 256-pixel fine slices achieving an optimal balance between accuracy and efficiency. As the first open-source TOD slicing framework on Hugging Face Apps and Google Colab, GOIS delivers real-time inference, open-source code, and live demonstrations, establishing itself as a breakthrough in object detection. The code and results are publicly available at https://github.com/MMUZAMMUL/GOIS with a live demoe at https://youtu.be/ukWUfXBFZ5I.}
}
📥 Quick Start
| Step | Command |
|----------|------------|
| 1️⃣ Clone Repo | git clone https://github.com/MMUZAMMUL/GOIS.git && cd GOIS |
| 2️⃣ Download Data | Follow Dataset Instructions or Download 15% Dataset |
| 3️⃣ Download Models | cd Models && python download_models.py |
| 4️⃣ Generate Ground Truth | python scripts/generate_ground_truth.py --annotations_folder "<annotations_path>" --images_folder "<images_path>" --output_coco_path "./data/ground_truth/ground_truth_coco.json" |
| 5️⃣ Full Inference (FI-Det) | python scripts/full_inference.py --images_folder "<path>" --model_path "Models/yolo11n.pt" --model_type "YOLO" --output_base_path "./data/FI_Predictions" |
| 6️⃣ GOIS Inference | python scripts/gois_inference.py --images_folder "<path>" --model_path "Models/yolo11n.pt" --model_type "YOLO" --output_base_path "./data/gois_Predictions" |
| 7️⃣ Evaluate FI-Det | python scripts/evaluate_prediction.py --ground_truth_path "./data/ground_truth/ground_truth_coco.json" --predictions_path "./data/FI_Predictions/full_inference.json" --iou_type bbox |
| 8️⃣ Evaluate GOIS-Det | python scripts/evaluate_prediction.py --ground_truth_path "./data/ground_truth/ground_truth_coco.json" --predictions_path "./data/gois_Predictions/gois_inference.json" --iou_type bbox |
| 9️⃣ Compare Results | python scripts/calculate_results.py --ground_truth_path "./data/ground_truth/ground_truth_coco.json" --full_inference_path "./data/FI_Predictions/full_inference.json" --gois_inference_path "./data/gois_Predictions/gois_inference.json" |
| 🔟 Upscale Metrics | python scripts/evaluate_upscaling.py --ground_truth_path "./data/ground_truth/ground_truth_coco.json" --full_inference_path "./data/FI_Predictions/full_inference.json" --gois_inference_path "./data/gois_Predictions/gois_inference.json" |
📊 Test GOIS Benchmarks & Gradio Live Deployement
📂 GOIS Benchmarks Repository
🎥 Watch Live Demo (YouTube) | 🎥 Watch Live Demo (Bilibili)
🔑 MIT License - Study & Educational Use Only
📧 Contact: Author Email
🚀 GOIS Live Deployed Applications on Gradio ✅
Explore the GOIS-Det vs. FI-Det benchmark results through live interactive applications on Gradio. These applications provide detailed comparisons using graphs, tables, and output images, demonstrating the effectiveness of GOIS-Det in tiny object detection.
🔥 Live Benchmark Tests different categories
| Test Function | Description | Live Test |
|-------------------|---------------|--------------|
| 1️⃣ Single Image Analysis (GOIS-Det vs. FI-Det) | Perform a single image test to visualize graphs and results, comparing FI-Det vs. GOIS-Det. View detection metrics such as the number of detections and class diversity. Outputs include pie charts, bar charts, and two comparative images that highlight the significance of GOIS. | 
📌 Instructions:
- Click on any "Open in Colab" button above to launch the interactive notebook.
- Follow the instructions in the notebook to test GOIS-Det vs. FI-Det.
- Evaluate detection performance using provided visualizations and metrics.
🚀 GOIS Live Deployed Applications on Hugging Face ✅ <img src="https://huggingface.co/front/assets/huggingface_logo-noborder.svg" alt="Hugging Face" width="32">
Experience Guided Object Inference Slicing (GOIS) across images, videos, and live cameras with configurable parameters. Evaluate real-time small object detection and compare against full-image inference (FI-Det).
📂 Compatible Datasets: VisDrone, UAV Surveillance (100-150ft), Pedestrian & Tiny Object Detection, Geo-Sciences
🖥️ Applied Models: YOLO11, YOLOv10, YOLOv9, YOLOv8, YOLOv6, YOLOv5, RT-DETR-L, YOLOv8s-Worldv2
GOIS incorporates a two-stage hierarchical slicing strategy, dynamically adjusting coarse-to-fine slicing and overlap rates to optimize tiny object detection while reducing false positives. These live applications allow users to test GOIS against full-image inference, analyze
