Fig. 1. Principle diagram of dual-camera defocused imaging

Fig. 2. Faster-RCNN network architecture^[19]

Fig. 3. VGG16 network architecture^[20]

Fig. 4. Flow chart of simultaneous particle size and position prediction based on convolution neural network

Fig. 5. Schematic diagram of dual-camera system and calibration test bench^[17]. (a) Schematic diagram of dual-camera system; (b) calibration test bench

Fig. 6. Partial dot images after cropping

Fig. 7. Comparison of particle size and depth errors under different particle sizes. (a)(b) particle size error; (c)(d) depth error

Fig. 8. Size prediction error of standard particles in circulating sample cell