Systems, methods and devices for monitoring betting activities

A machine-learning based method for machine vision bet recognition by visually capturing chip placement activities, the method comprising:

• capturing, using a camera or sensor, image data corresponding to one or more chips positioned in at least one chip placement area captured across a plurality of channels including at least a red channel, a green channel, and a blue channel;
• establishing a depth information channel providing depth values estimated using at least one of:
• Doppler radar measurements, parallax measurements, infrared thermography, shadow measurements, light intensity measurements, relative size measurements, and illumination grid measurements;
• processing the captured image data by generating one or more representative images for each chip of the one or more chips for distinguishing different chip types, at least one of the one or more representative images generated from at least two channels of the plurality of channels including the depth information channel and at least one of the red channel, the green channel, and the blue channel;
• assigning, using a trained neural network, a chip type to each of the one or more chips by classifying each chip of the one or more chips by identifying a most similar corresponding reference classification to the representative image from a set of reference classifications, the trained neural network trained during a calibration process using real-world images of reference chips to train a base set of features for pattern recognition to dynamically build the set of reference classifications;
• and
• determining a total chip value or a change in the total chip value of the one or more chips in the at least one chip placement area based on the assigned chip types.

A non-transitory computer readable medium storing machine interpretable instructions, which when executed by a processor, perform a machine vision bet recognition method, the method comprising:

• capturing, using a camera or sensor, image data corresponding to one or more chips positioned in at least one chip placement area captured across a plurality of channels including at least a red channel, a green channel, and a blue channel;
• establishing a depth information channel providing depth values estimated using at least one of:
• Doppler radar measurements, parallax measurements, infrared thermography, shadow measurements, light intensity measurements, relative size measurements, and illumination grid measurements;
• processing the captured image data by generating one or more representative images for each chip of the one or more chips for distinguishing different chip types, at least one of the one or more representative images generated from at least two channels of the plurality of channels including the depth information channel and at least one of the red channel, the green channel, and the blue channel;
• assigning using a trained neural network, a chip type to each of the one or more chips by identifying a most similar corresponding reference classification to the representative image from a set of reference classifications, the trained neural network trained during a calibration process using real-world images of reference chips to train a base set of features for pattern recognition to dynamically build the set of reference classifications;
• and
• determining a total chip value or a change in the total chip value of the one or more chips in the at least one chip placement area based on the assigned chip types.