# Object Detection from Scratch with Deep Supervision

arXiv: https://arxiv.org/abs/1809.09294 · HTML: https://mlsanity.com/paper/1809.09294/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Sep 2018.

## Built on by (most cited first, 21 of 21)

- [YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors](https://mlsanity.com/paper/2207.02696/) (2022)
- [YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information](https://mlsanity.com/paper/2402.13616/) (2024)
- [TransVOD: End-to-End Video Object Detection with Spatial-Temporal Transformers](https://mlsanity.com/paper/2201.05047/) (2022)
- [Scaling Laws for Downstream Task Performance of Large Language Models](https://mlsanity.com/paper/2402.04177/) (2024)
- [Self-Adversarial Disentangling for Specific Domain Adaptation](https://mlsanity.com/paper/2108.03553/) (2021)
- [Proper Reuse of Image Classification Features Improves Object Detection](https://mlsanity.com/paper/2204.00484/) (2022)
- [MNet: Rethinking 2D/3D Networks for Anisotropic Medical Image Segmentation](https://mlsanity.com/paper/2205.04846/) (2022)
- [AWSnet: An Auto-weighted Supervision Attention Network for Myocardial Scar and Edema Segmentation in Multi-sequence Cardiac Magnetic Resonance Images](https://mlsanity.com/paper/2201.05344/) (2022)
- [A Comprehensive Review on Deep Supervision: Theories and Applications](https://mlsanity.com/paper/2207.02376/) (2022)
- [A Unified Light Framework for Real-time Fault Detection of Freight Train Images](https://mlsanity.com/paper/2102.00381/) (2021)
- [LocTex: Learning Data-Efficient Visual Representations from Localized Textual Supervision](https://mlsanity.com/paper/2108.11950/) (2021)
- [X-Learner: Learning Cross Sources and Tasks for Universal Visual Representation](https://mlsanity.com/paper/2203.08764/) (2022)
- [DPNet: Dual-Path Network for Real-time Object Detection with Lightweight Attention](https://mlsanity.com/paper/2209.13933/) (2022)
- [Artificial and beneficial -- Exploiting artificial images for aerial vehicle detection](https://mlsanity.com/paper/2104.03054/) (2021)
- [Automatic Semantic Segmentation of the Lumbar Spine: Clinical Applicability in a Multi-parametric and Multi-centre Study on Magnetic Resonance Images](https://mlsanity.com/paper/2111.08712/) (2021)
- [You Only Look Once at Anytime (AnytimeYOLO): Analysis and Optimization of Early-Exits for Object-Detection](https://mlsanity.com/paper/2503.17497/) (2025)
- [Data-Agnostic Augmentations for Unknown Variations: Out-of-Distribution Generalisation in MRI Segmentation](https://mlsanity.com/paper/2505.10223/) (2025)
- [A Comprehensive Review of AI-enabled Unmanned Aerial Vehicle: Trends, Vision , and Challenges](https://mlsanity.com/paper/2310.16360/) (2023)
- [Multi-channel Deep Supervision for Crowd Counting](https://mlsanity.com/paper/2103.09553/) (2021)
- [Exploring Camera Encoder Designs for Autonomous Driving Perception](https://mlsanity.com/paper/2407.07276/) (2024)
- [Guided Learning: Lubricating End-to-End Modeling for Multi-stage Decision-making](https://mlsanity.com/paper/2411.10496/) (2024)

## Abstract (arXiv)

We propose Deeply Supervised Object Detectors (DSOD), an object detection framework that can be trained from scratch. Recent advances in object detection heavily depend on the off-the-shelf models pre-trained on large-scale classification datasets like ImageNet and OpenImage. However, one problem is that adopting pre-trained models from classification to detection task may incur learning bias due to the different objective function and diverse distributions of object categories. Techniques like fine-tuning on detection task could alleviate this issue to some extent but are still not fundamental. Furthermore, transferring these pre-trained models across discrepant domains will be more difficult (e.g., from RGB to depth images). Thus, a better solution to handle these critical problems is to train object detectors from scratch, which motivates our proposed method. Previous efforts on this direction mainly failed by reasons of the limited training data and naive backbone network structures for object detection. In DSOD, we contribute a set of design principles for learning object detectors from scratch. One of the key principles is the deep supervision, enabled by layer-wise dense connections in both backbone networks and prediction layers, plays a critical role in learning good detectors from scratch. After involving several other principles, we build our DSOD based on the single-shot detection framework (SSD). We evaluate our method on PASCAL VOC 2007, 2012 and COCO datasets. DSOD achieves consistently better results than the state-of-the-art methods with much more compact models. Specifically, DSOD outperforms baseline method SSD on all three benchmarks, while requiring only 1/2 parameters. We also observe that DSOD can achieve comparable/slightly better results than Mask RCNN + FPN (under similar input size) with only 1/3 parameters, using no extra data or pre-trained models.