Abstract: A system and method of fine-grained image category discovery with few human annotations includes a camera and a trained machine learning model, which predicts a label for an object in a captured image and outputs the predicted label. The machine learning model is trained by contrastive affinity learning, including retrieving images having an object, a warm-up stage in which semi-supervised contrastive learning is performed based on projected features of a class token and an ensembled prompt, respectively. In a contrastive affinity learning stage, a student model and an exponentially moving averaged teacher model are forwarded with different augmented views of the retrieved images. Teacher embeddings are enqueued into a token-specific memory. A semi-supervised contrastive loss is computed on a current batch and a contrastive affinity learning loss for student embeddings and the teacher embeddings with pseudo-labels from a affinity graph dynamically generated by semi-supervised affinity generation.

Abstract: A system and method of fine-grained image category discovery with few human annotations includes a camera and a trained machine learning model, which predicts a label for an object in a captured image and outputs the predicted label. The machine learning model is trained by contrastive affinity learning, including retrieving images having an object, a warm-up stage in which semi-supervised contrastive learning is performed based on projected features of a class token and an ensembled prompt, respectively. In a contrastive affinity learning stage, a student model and an exponentially moving averaged teacher model are forwarded with different augmented views of the retrieved images. Teacher embeddings are enqueued into a token-specific memory. A semi-supervised contrastive loss is computed on a current batch and a contrastive affinity learning loss for student embeddings and the teacher embeddings with pseudo-labels from a affinity graph dynamically generated by semi-supervised affinity generation.

Abstract: An image registration method is disclosed for processing a distorted image into a registered image that is aligned with reference to an original image. Distortions from the original image may include scaling, rotation, and noise. The method is based on correlating Radon transforms of both images to determine the rotation angle, and the scaling factor is determined by dividing averages of the overall luminance of each image on the assumption that any added noise will cancel. The Fast Fourier Transform (FFT) is used to estimate global spatial shifts. In one embodiment, the distorted image is first scaled to the size of the original image before being rotated. In another embodiment, the original image is first scaled to the size of the distorted image before rotating the distorted image, and finally scaling it to match the original image. A corresponding system for image registration is also provided.

Abstract: An image registration method is disclosed for processing a distorted image into a registered image that is aligned with reference to an original image. Distortions from the original image may include scaling, rotation, and noise. The method is based on correlating Radon transforms of both images to determine the rotation angle, and the scaling factor is determined by dividing averages of the overall luminance of each image on the assumption that any added noise will cancel. The Fast Fourier Transform (FFT) is used to estimate global spatial shifts. In one embodiment, the distorted image is first scaled to the size of the original image before being rotated. In another embodiment, the original image is first scaled to the size of the distorted image before rotating the distorted image, and finally scaling it to match the original image. A corresponding system for image registration is also provided.