Image-free single-pixel segmentation

The existing segmentation techniques require high-fidelity images as input to perform semantic segmentation. Since the results contain most of edge information that is much less than acquired images, throughput gap leads both hardware and software waste. In this paper, we report an image-free single-pixel technique. technique combines structured illumination detection together, efficiently sample multiplex scene’s into compressed one-dimensional measurements. patterns are optimized together with subsequent reconstruction neural network, which directly infers maps from end-to-end encoding-and-decoding learning framework enables corresponding provides high acquisition efficiency. Both simulation experimental validate accurate can be achieved using two-order-of-magnitude data. When sampling ratio 1%, Dice coefficient reaches above 80% pixel accuracy 96%. We envision widely applied in various resource-limited platforms such Unmanned Aerial Vehicle (UAV) autonomous vehicle real-time sensing. • An method 1-D improved system’s Experimental show at 1% ratio.