Neural network based on temporal attention for video compression

Systems, methods, and instrumentalities are disclosed for video encoding and/or video decoding using artificial neural networks (e.g., convolutional neural networks or recurrent neural networks), attention, and/or attention with spatial attributes. For example, an apparatus may be configured to perform one or more of: obtaining a context block, a current block, and a potential vector associated with the context block; performing at least one convolution on the context block, the reference block and the potential vector; generating motion stream data associated with the current block based on the at least one convolution; or generating a bitstream comprising an indication of the motion stream data. The motion stream data may be quantized. The generated bitstream may include an indication of quantized motion stream data. 公开了用于使用人工神经网络(例如，卷积神经网络或循环神经网络)、注意力和/或具有空间属性的注意力进行视频编码和/或视频解码的系统、方法和手段。例如，一种装置，该装置可被配置为执行以下中的一者或多者：获得上下文块、当前块和与该上下文块相关联的潜在向量；对该上下文块、参考块和该潜在向量执行至少一次卷积；基于该至少一次卷积生成与该当前块相关联的运动流数据；或者生成包括对该运动流数据的指示的