Deep Face Detector Adaptation Without Negative Transfer or Catastrophic Forgetting

Jamal, Muhammad Abdullah; Li, Haoxiang; Gong, Boqing

doi:10.1109/cvpr.2018.00588

Cited by 10 publications

(9 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The FDDB dataset [24] is comprised of 5,171 annotated faces in a set of 2,845 images taken from a subset of the Face in the Wild dataset. The images and the annotation style of FDDB have a significant domain shift from WIDER Face, which are discussed in Jamal et al [1]. Fig.…”

Section: Discussionmentioning

confidence: 99%

Unsupervised Hard Example Mining from Videos for Improved Object Detection

Jin

RoyChowdhury

Jiang

et al. 2018

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Important gains have recently been obtained in object detection by using training objectives that focus on hard negative examples, i.e., negative examples that are currently rated as positive or ambiguous by the detector. These examples can strongly influence parameters when the network is trained to correct them. Unfortunately, they are often sparse in the training data, and are expensive to obtain. In this work, we show how large numbers of hard negatives can be obtained automatically by analyzing the output of a trained detector on video sequences. In particular, detections that are isolated in time, i.e., that have no associated preceding or following detections, are likely to be hard negatives. We describe simple procedures for mining large numbers of such hard negatives (and also hard positives) from unlabeled video data. Our experiments show that retraining detectors on these automatically obtained examples often significantly improves performance. We present experiments on multiple architectures and multiple data sets, including face detection, pedestrian detection and other object categories.

show abstract

Section: Discussionmentioning

confidence: 99%

Unsupervised Hard Example Mining from Videos for Improved Object Detection

Jin

RoyChowdhury

Jiang

et al. 2018

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Despite the common motivation, our work differs on two major points -(a) we show the usefulness of combining both hard and easy examples from the target domain when re-training the baseline model, and (b) using the knowledge distillation loss to counter the effect of label noise. Jamal et al [1] address the domain shift between various face detection datasets by recalibrating the final classification layer of face detectors using a residual-style layer in a low-shot learning setting. Two recent methods [23,8] for domain-adaptive object detection are particularly relevant to our problem.…”

Section: Related Workmentioning

confidence: 99%

Automatic Adaptation of Object Detectors to New Domains Using Self-Training

RoyChowdhury

Chakrabarty

Singh

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

139

View full text Add to dashboard Cite

This work addresses the unsupervised adaptation of an existing object detector to a new target domain. We assume that a large number of unlabeled videos from this domain are readily available. We automatically obtain labels on the target data by using high-confidence detections from the existing detector, augmented with hard (misclassified) examples acquired by exploiting temporal cues using a tracker. These automatically-obtained labels are then used for re-training the original model. A modified knowledge distillation loss is proposed, and we investigate several ways of assigning soft-labels to the training examples from the target domain. Our approach is empirically evaluated on challenging face and pedestrian detection tasks: a face detector trained on WIDER-Face, which consists of highquality images crawled from the web, is adapted to a largescale surveillance data set; a pedestrian detector trained on clear, daytime images from the BDD-100K driving data set is adapted to all other scenarios such as rainy, foggy, nighttime. Our results demonstrate the usefulness of incorporating hard examples obtained from tracking, the advantage of using soft-labels via distillation loss versus hard-labels, and show promising performance as a simple method for unsupervised domain adaptation of object detectors, with minimal dependence on hyper-parameters. Code and models are available at

show abstract

“…The overwhelming majority of research on CV considered the development of robust algorithms to improve accuracy in static image datasets [4,5] (and references within), fewer dealt with videos [13], and even fewer contributions addressed system design aspects [14].…”

Section: Relationship To Prior Workmentioning

confidence: 99%

“…An example of such tests is face detection, which can result in either detecting a face or not. Many recent literatures also use recall to report system sensitivity, for instance recent literature like [5], also use recall to measure the performance of face detector adaptation to training with different datasets. We use the recall error (denoted by E) to measure the system sensitivity.…”

Section: Model Developmentmentioning

confidence: 99%

“…Therefore, these adaptations lead to various tradeoffs involving the accuracy and the aforementioned constraints. The overwhelming majority of studies on CV focused on the development of robust algorithms to improve the accuracy in primarily image datasets, through statistical [1] (and references within) and more recently deep learning approaches [2,3,4,5] (and references within).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

QRMODA and BRMODA: Novel Models for Face Recognition Accuracy in Computer Vision Systems with Adapted Video Streams

Hamandi,

Sarhan

2019

Preprint

View full text Add to dashboard Cite

A major challenge facing Computer Vision systems is providing the ability to accurately detect threats and recognize subjects and/or objects under dynamically changing network conditions. We propose two novel models that characterize the face recognition accuracy in terms of video encoding parameters. Specifically, we model the accuracy in terms of video resolution, quantization, and actual bit rate. We validate the models using two distinct video datasets and a large image dataset by conducting 1, 668 experiments that involve simultaneously varying combinations of encoding parameters. We show that both models hold true for the deep learning and statistical based face recognition. Furthermore, we show that the models can be used to capture different accuracy metrics, specifically the recall, precision, and F1-score. Ultimately, we provide meaningful insights on the factors affecting the constants of each proposed model.

show abstract

Deep Face Detector Adaptation Without Negative Transfer or Catastrophic Forgetting

Cited by 10 publications

References 28 publications

Unsupervised Hard Example Mining from Videos for Improved Object Detection

Unsupervised Hard Example Mining from Videos for Improved Object Detection

Automatic Adaptation of Object Detectors to New Domains Using Self-Training

QRMODA and BRMODA: Novel Models for Face Recognition Accuracy in Computer Vision Systems with Adapted Video Streams

Contact Info

Product

Resources

About