Method and system for performing object detection using a convolutional neural network

A camera comprising:

• a housing through which extends an aperture that permits light to enter the housing;
• an image sensor contained within the housing to receive the light that has entered the housing through the aperture;
• a processor communicatively coupled to the image sensor;
• and
• a memory communicatively coupled to the processor, wherein the memory has stored thereon computer program code that is executable by the processor and that, when executed by the processor, causes the processor to perform a method comprising:
• obtaining an image;
• generating at least one feature pyramid by processing the image using a convolutional neural network (CNN);
• splitting an enhanced feature map comprising part of the at least one feature pyramid into channel feature maps of different resolutions, wherein the CNN comprises at least one feature map that is shallower than the enhanced feature map, and wherein each of the channel feature maps corresponds to channels of the enhanced feature map;
• and
• performing object detection on the channel feature maps.

A non-transitory computer readable medium having stored thereon computer program code that is executable by a processor and that, when executed by the processor, causes the processor to perform a method comprising:

• obtaining an image;
• generating a first feature pyramid by processing the image using a convolutional neural network (CNN);
• generating a second feature pyramid from the first feature pyramid using the CNN, wherein the second feature pyramid comprises an enhanced feature map generated by combining an upsampled feature map and a feature map of the first feature pyramid that has a corresponding or lower resolution of a resolution of the enhanced feature map, and wherein the upsampled feature map is generated by upsampling a feature map of the second feature pyramid that is at a shallower position in the CNN than the enhanced feature map;
• splitting the enhanced feature map into channel feature maps of different resolutions, wherein each of the channel feature maps corresponds to channels of the enhanced feature map;
• and
• performing object detection on the channel feature maps.

A non-transitory computer readable medium having stored thereon computer program code that is executable by a processor and that, when executed by the processor, causes the processor to perform a method comprising:

• obtaining an image;
• generating at least one feature pyramid by processing the image using a convolutional neural network (CNN);
• splitting an enhanced feature map comprising part of the at least one feature pyramid into channel feature maps of different resolutions, wherein the CNN comprises at least one feature map that is shallower than the enhanced feature map, and wherein each of the channel feature maps corresponds to channels of the enhanced feature map;
• and
• performing object detection on the channel feature maps.