--- license: mit language: - en --- # Chromatic spikes ![Chromatic spikes illustration](https://cdn-uploads.huggingface.co/production/uploads/65b247e77cd0069ad6f176fa/cJ7f8uxwqWO9HSP2CftKN.png) This is the dataset for **Real-data-driven 2000 FPS Color Video from Mosaicked Chromatic Spikes**. ## Dataset Details ### Description This is the public dataset containing more than 100 chromatic spike streams, captured by a spike camera equipped with Bayer-pattern color filter array. Each stream can be represented as an uint8 array with shape [T, H, W], where T is the number of spike frames, H and W are height and width (1000 and 1000 for this dataset). Spike streams are saved with [**tensorstore**](https://google.github.io/tensorstore/), which is a library for efficiently reading and writing large multi-dimensional arrays. To load the spike streams, please first extract tar files and refer to the following example script: ```python import numpy as np import tensorstore as ts dataset_dir = "tensorstore/20230922113135162" dataset = ts.open( { "driver": "zarr3", "kvstore": { "driver": "file", "path": dataset_dir, }, } ).result() print(dataset.shape) # => (20000, 1000, 1000) print(dataset.dtype) # => dtype("uint8") print(np.unique(dataset[:100].read().result())) # => [0 1] ``` ### Sources - **Repository:** [Comming soon] - **Paper:** [Real-data-driven 2000 FPS Color Video from Mosaicked Chromatic Spikes] ## Structure The dataset contains more than 100 chromatic spike streams, each is saved as a tensorstore array and further archived to a tar file. The tar files are placed in `/tensorstore/`. ## Citation **BibTeX:** [Comming soon] ## Authors [yousiki](https://www.siki.moe/about/) ## Contact [me at siki dot moe](mailto:me@siki.moe)