Look One and More: Distilling Hybrid Order Relational Knowledge for Cross-Resolution Image Recognition

Abstract

In spite of great success in many image recognition tasks achieved by recent deep models, directly applying them to recognize low-resolution images may suffer from low accuracy due to the missing of informative details during resolution degradation. However, these images are still recognizable for subjects who are familiar with the corresponding high-resolution ones. Inspired by that, we propose a teacher-student learning approach to facilitate low-resolution image recognition via hybrid order relational knowledge distillation. The approach refers to three streams: the teacher stream is pretrained to recognize high-resolution images in high accuracy, the student stream is learned to identify low-resolution images by mimicking the teacher's behaviors, and the extra assistant stream is introduced as bridge to help knowledge transfer across the teacher to the student. To extract sufficient knowledge for reducing the loss in accuracy, the learning of student is supervised with multiple losses, which preserves the similarities in various order relational structures. In this way, the capability of recovering missing details of familiar low-resolution images can be effectively enhanced, leading to a better knowledge transfer. Extensive experiments on metric learning, low-resolution image classification and low-resolution face recognition tasks show the effectiveness of our approach, while taking reduced models.

Student Network Learning via Evolutionary Knowledge Distillation

“…The dataset used is down-sampled to reduce the resolution to 16×16, and the classification loss is classic cross-entropy loss.The experimental results are shown in Fig. 6, our approach shows good adaptability on more challenging low-resolution image classification task, which is 1.31% higher than the current best method HORKD [64]. In addition, HORKD needs to consume more computing resources due to the introduction of the assistant network.…”

“…Moreover, their models have more parameters, which will lead to a significant increase in the computing cost for inference. It is worth mentioning that whether it is VLRR [65], SKD [62], HORKD [64] or LRFRW [71] in the distillation process, high-resolution images corresponding to low-resolution faces are necessary to provide more information, but such highresolution images are not always easy to obtain, our approach only uses low-resolution face images for training, which adopts real-world application scenarios.…”

“…ResNet50 and ResNet18 are chosen as teacher and student, respectively. We compare the performance of student models with several typical distillation methods include KD [9], FitNet [12] and CRD [64]. The percentages of retained samples are 100%, 75%, 50% and 25%.…”

Deep Rival Penalized Competitive Learning for Low-resolution Face Recognition

“…[18] proposed that the direct transfer from private HR to wild LR may be difficult, and used public HR and LR as a bridge to distill and compress knowledge. In order to fully transfer HR recognition knowledge to LR network, [20] not only focuses on the first-order knowledge learning between points, but also considers high-order distillation. The knowledge of various order relationships is extracted from the teacher network as a supervision signal for student network.…”

“…There are also works that use knowledge distillation [16,17,18,19,20] to make an LR network mimic a HR network which is obtained under a rich HR training set. With paired LR and HR images of the same identity, the LR network is supervised by both class label and the soft target of HR network, corresponding to classification loss and distillation loss, respectively.…”