Deep learning system for cuboid detection
Systems and methods for cuboid detection and keypoint localization in images are disclosed. In one aspect, a deep cuboid detector can be used for simultaneous cuboid detection and keypoint localization in monocular images. The deep cuboid detector can include a plurality of convolutional layers and non-convolutional layers of a trained convolution neural network for determining a convolutional feature map from an input image. A region proposal network of the deep cuboid detector can determine a bounding box surrounding a cuboid in the image using the convolutional feature map. The pooling layer and regressor layers of the deep cuboid detector can implement iterative feature pooling for determining a refined bounding box and a parameterized representation of the cuboid.
1. A system comprising:
non-transitory memory configured to store executable instructions, and
one or more hardware processors in communication with the non-transitory memory, the one or more hardware processors programmed by the executable instructions to:
access an image;
generate a convolutional feature map based at least in part on a first convolutional neural network (CNN) and the image;
determine at least one region of interest (RoI) based at least in part on a region proposal network (RPN), wherein the RoI comprises a cuboid at a cuboid image location of the image;
determine a submap of the convolutional feature map corresponding to the RoI based at least in part on a pooling layer and the cuboid image location; and
determine a refined RoI at a refined cuboid image location and a representation of the cuboid based at least in part on at least one regressor layer and the submap of the convolutional feature map corresponding to the RoI.
2. The system of claim 1 , wherein the one or more hardware processors are further programmed to:
determine a refined submap of the convolutional feature map corresponding to the refined RoI based at least in part on the refined cuboid image location;
determine a further refined RoI at a further refined cuboid image location and a further defined representation of the cuboid based at least in part on the pooling layer, the at least one regressor layer, and the refined submap of the convolutional feature map corresponding to the refined RoI.
3. The system of claim 1 , wherein the cuboid image location or the refined cuboid image location are represented as a two-dimensional (2D) bounding box.
4. The system of claim 1 , wherein the at least one regressor layer comprises a fully connected layer, a non-fully connected layer, or any combination thereof.
5. The system of claim 1 , wherein the RPN comprises a deep neural network (DNN).
6. The system of claim 1 , wherein the representation of the cuboid comprises a parameterized representation of the cuboid.
7. The system of claim 6 , wherein the parameterized representation of the cuboid comprises locations of a plurality of keypoints of the cuboid in the image.
8. The system of claim 7 , wherein the parameterized representation comprises normalized offsets of the plurality of keypoints of the cuboid from the center of the image.
9. The system of claim 6 , wherein the parameterized representation of the cuboid comprises a vanishing point parameterization.
10. The system of claim 1 , wherein the one or more hardware processors are further programmed to:
interact with a user based on the refined RoI at the refined cuboid image location and the representation of the cuboid.
11. The system of claim 10 , wherein the cuboid corresponds to a stationary box, and wherein to interact with the user, the one or more hardware processors are further programmed to:
generate character animation in relation to the stationary box based on the refined image location of the cuboid and the representation of the cuboid.
12. The system of claim 10 , wherein the cuboid corresponds to a hand-held cuboid, and wherein to interact with the user of the system, the one or more hardware processors are further programmed to:
determine a pose of the cuboid using the representation of the cuboid; and
interact with the user of the system based on the pose of the cuboid.
13. The system of claim 10 , wherein the cuboid corresponds to a rare object not recognizable by a second CNN, and wherein to interact with the user of the system, the one or more hardware processors are further programmed to:
provide the user with a notification that the rare object not recognizable by the second CNN is detected.
14. The system of claim 1 , wherein the cuboid corresponds to a marker, and wherein the one or more hardware processors are further programmed to:
perform simultaneous location and mapping (SLAM) based on the refined RoI at the refined cuboid image location and the representation of the cuboid.
15. A method for cuboid detection and keypoint localization comprising, under control of a one or more hardware processors:
accessing an image;
generating a convolutional feature map based at least in part on a first convolutional neural network (CNN) and the image;
determining at least one region of interest (RoI) based at least in part on a region proposal network (RPN) of a cuboid detector, wherein the RoI includes a cuboid at a cuboid image location of the image;
determining a submap of the convolutional feature map corresponding to the RoI based at least in part on a pooling layer of the cuboid detector and the cuboid image location; and
determining a refined RoI at a refined cuboid image location and the representation of the cuboid based at least in part on at least one regressor layer of the cuboid detector and the submap of the convolutional feature map corresponding to the RoI.
16. The method of claim 15 , further comprising:
determining a refined submap of the convolutional feature map corresponding to the refined RoI based at least in part on the refined cuboid image location;
determining a further refined RoI at a further refined cuboid image location and a further defined representation of the cuboid based at least in part on the pooling layer, the at least one regressor layer, and the refined submap of the convolutional feature map corresponding to the refined RoI.
17. The method of claim 15 , wherein the cuboid image location or the refined cuboid image location is represented as a two-dimensional (2D) bounding box.
18. The method of claim 15 , wherein the non-convolutional layers of the first CNN comprises a normalization layer, a brightness normalization layer, a batch normalization layer, a rectified linear layer, an upsampling layer, a concatenation layer, a pooling layer, a softsign layer, or any combination thereof.
19. The method of claim 15 , wherein the representation of the cuboid comprises a parameterized representation of the cuboid comprising locations of a plurality of keypoints of the cuboid in the image.
20. The method of claim 15 , wherein the at least one regressor layer comprises a fully connected layer, a non-fully connected layer, or any combination thereof.