English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Obtaining the ground truth labels from a video is challenging since themanual annotation of pixel-wise flow labels is prohibitively expensive andlaborious. Besides, existing approaches try to adapt the trained model onsynthetic datasets to authentic videos, which inevitably suffers from domaindiscrepancy and hinders the performance for real-world applications. To solvethese problems, we propose RealFlow, an Expectation-Maximization basedframework that can create large-scale optical flow datasets directly from anyunlabeled realistic videos. Specifically, we first estimate optical flowbetween a pair of video frames, and then synthesize a new image from this pairbased on the predicted flow. Thus the new image pairs and their correspondingflows can be regarded as a new training set. Besides, we design a RealisticImage Pair Rendering (RIPR) module that adopts softmax splatting andbi-directional hole filling techniques to alleviate the artifacts of the imagesynthesis. In the E-step, RIPR renders new images to create a large quantity oftraining data. In the M-step, we utilize the generated training data to trainan optical flow network, which can be used to estimate optical flows in thenext E-step. During the iterative learning steps, the capability of the flownetwork is gradually improved, so is the accuracy of the flow, as well as thequality of the synthesized dataset. Experimental results show that RealFlowoutperforms previous dataset generation methods by a considerably large margin.Moreover, based on the generated dataset, our approach achievesstate-of-the-art performance on two standard benchmarks compared with bothsupervised and unsupervised optical flow methods. Our code and dataset areavailable at https://github.com/megvii-research/RealFlow

RealFlow: EM-based Realistic Optical Flow Dataset Generation from Videos