Revisiting Knowledge Transfer for Training Object Class Detectors

By Jasper R. R. Uijlings, Stefan Popov, Vittorio Ferrari. 2017

73 citations recorded by the source corpus.

We propose to revisit knowledge transfer for training object detectors on target classes from weakly supervised training images, helped by a set of source classes with bounding-box annotations. We present a unified knowledge transfer framework based on training a single neural network multi-class object detector over all source classes, organized in a semantic hierarchy. This generates proposals with scores at multiple levels in the hierarchy, which we use to explore knowledge transfer over a broad range of generality, ranging from class-specific (bycicle to motorbike) to class-generic (objectness to any class). Experiments on the 200 object classes in the ILSVRC 2013 detection dataset show that our technique (1) leads to much better performance on the target classes (70.3% CorLoc, 36.9% mAP) than a weakly supervised baseline which uses manually engineered objectness [11] (50.5% CorLoc, 25.4% mAP). (2) delivers target object detectors reaching 80% of the mAP of their fully supervised counterparts. (3) outperforms the best reported transfer learning results on this dataset (+41% CorLoc and +3% mAP over [18, 46], +16.2% mAP over [32]). Moreover, we also carry out several across-dataset knowledge transfer experiments [27, 24, 35] and find that (4) our technique outperforms the weakly supervised baseline in all dataset pairs by 1.5 × - 1.9×, establishing its general applicability.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Mohit Bansal
2. Quoc V. Le
3. Jianfeng Gao
4. Ishan Misra
5. Trevor Darrell
6. Armand Joulin
7. Abhinav Kumar Gupta
8. Caiming Xiong

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Jitendra Malik
3. Dacheng Tao
4. Xiaogang Wang
5. Trevor Darrell
6. Ming-Hsuan Yang
7. Andrew Zisserman
8. Cordelia Schmid

Citations in the corpus

• Learning Open-World Object Proposals without Learning to Classify (2021, methodology): “Approaches under semi-supervised setup [27,44,49,50,55,58,73] employ the weak image-level labels for novel classes as well as boxlevel labels for base classes.”
• The Open Images Dataset V4 (2018, methodology): “We train our initial detector using the weakly-supervised technique with knowledge transfer described in (Uijlings et al., 2018).”
• Relational Knowledge Distillation (2019, methodology): “With growing interests in KD, task-specific KD methods have been proposed for object detection [5, 6, 37], face model compression [24], and image retrieval and Re-ID [7].”
• Open-Vocabulary Object Detection Using Captions (2020, background): “…to LSDA [18], which learns a transformation from classifier weights into detector weights, its extension [36] to utilize semantic class relationships (LSDA+), and a more recent work [37] which uses multiple-instance learning on a region proposal network that is pretrained on base classes (MIL+RPN).”
• Weakly Supervised Object Localization and Detection: A Survey (2021, unknown): “Recently, [151] revisits knowledge transfer for training the weakly supervised object detector.”
• Relationship Matters: Relation Guided Knowledge Transfer for Incremental Learning of Object Detectors (2020, unknown): “With growing interests in knowledge distillation, task-specific KD methods have also been proposed for object detection [39, 4, 40, 29] but with fixed number of classes.”
• What leads to generalization of object proposals? (2020, background): “[35] tried to transfer information from coarse source classes to fine-grained target classes that share similar localization properties.”
• Detecting 11K Classes: Large Scale Object Detection Without Fine-Grained Bounding Boxes (2019, methodology): “[35] proposes a semiMIL method on disjoint set semi-supervised detection, which achieves better performance than [32].”
• Webly Supervised Concept Expansion for General Purpose Vision Models (2022, methodology): “[74] use Multiple Instance Learning to pseudo-label data and train a large vocabulary detector.”
• Detecting Twenty-thousand Classes using Image-level Supervision (2022, background): “DLWL [39] combines self-training and clustering-based WSOD [53], and again assigns image labels to max-scored proposals.”
• R-FCN-3000 at 30fps: Decoupling Detection and Classification (2017, methodology): “Contemporary work [38] (KDT) also employs a knowledge transfer based approach, albeit with a modern Inception-ResNet-v2 based Faster-RCNN detector.”