Sports-1M is a dataset consists of 1 million youtube videos annotated with 487 classes. the classes are arranged in a manually curated text that contains internal nodes of sports related taxonomy. There are 1000-3000 videos per class, some of the videos re annotated with more than one class.

University of Michigan - Ann Arbor

Google

- As for the CNN Models (single frame/ early fusion/ late fusion/ slow fusion), they learn well compared to the baseline model. The single-frame model already displays strong performance.
- Compared to the wide gap relative to the feature-based baseline, the variation among different CNN architectures are comparatively insignificant.
- Multiresolution CNN architectures are 2-4 times faster due to reduced dimensions from input layer.


Quantitative Findings of the Sports-1M video classification experiments using CNNs and feature histogram baseline models

Karpathy, A., Toderici, G., Shetty, S., Leung, T., Sukthankar, R., & Fei-Fei, L. (2014). Large-scale video classification with convolutional neural networks. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition (pp. 1725-1732). https://doi.org/10.1109/cvpr.2014.223

Large-Scale Video Classification with Convolutional Neural Networks

Sports-1M Dataset

Several types of features, such as  Hue-Saturation, Color moments, number of faces detected, and sparse local interest points, were extracted at all frames of videos and discretized using k-means vector quantization and accumulated into histograms. All histograms are concatenated into a 25,000 dimensional video-level feature vector, connectd with multilayer neural network with Relu followed by a Softmax classifier.

Learn Before

Related