Method and system for unsupervised word image clustering

A system for unsupervised word image clustering said system comprising

• at least one image capture device operatively coupled to the system,
• a processor, an interface, and memory, wherein the memory, store instructions for execution by the processors to:
• capture one or more image using at least one image capture device, wherein at least one of the one or more image comprises at least one word image;
• extract one or more feature vector for each of the at least one word image using an untrained convolution neural network architecture, wherein extraction is performed by:
• a first convolution to the at least one input image using a first plurality of filters, wherein the first plurality of filters are initialized by random filter based deep learning techniques using Gaussian random variable with zero mean and unit standard deviation, and wherein the plurality of filters are constrained to sum to zero;
• apply a Rectified Linear Unit (ReLU) non-linearity to a first plurality of feature maps, wherein the first plurality of feature maps is generated as output of the first convolution;
• apply a first sub-sampling to increase the position invariance of the first plurality of feature maps, wherein sub-sampling comprises applying a pooling operation on non overlapping segments of the first plurality of feature maps after application of the ReLU non-linearity;
• apply a second convolution on the output of the first sub-sampling using a second plurality of filters wherein the second plurality of filters are initialized by random filter based deep learning techniques using Gaussian random variable with zero mean and unit standard deviation, and wherein the plurality of filters are constrained to sum to zero;
• apply a Rectified Linear Unit (ReLU) non-linearity to a second plurality of feature maps, wherein the second plurality of feature maps are generated as output of the second convolution;
• apply a second sub-sampling to increase the position invariance of the second plurality of feature maps, wherein sub-sampling comprises applying a pooling operation on non-overlapping segments of the second plurality of feature maps after application of the ReLU non linearity;
• and
• combine the plurality of feature maps generated at the output of the second sub sampling for extracting one or more feature vector;
• and
• cluster, the one or more word images, wherein clustering is based on the one or more feature vector, the clustering comprises:
• clustering word images based on a similarity value across a first stage cluster, wherein the similarity value is evaluated by comparing the extracted one or more feature vector with a pre-defined first similarity threshold value (Tstrong);
• and
• clustering word images based on a similarity value across a second stage cluster, wherein the similarity value is evaluated by comparing the extracted one or more feature vector with a pre-defined second similarity threshold value (Tsoft),
• wherein the first similarity threshold value (Tstrong) is higher than the second similarity threshold value (Tsoft),
• wherein the clustered word images are used for performing one or more functions, wherein the one or more functions comprises indexing, transcription, and optical character recognition.

A non-transitory computer readable medium storing instructions which when executed by a possessor on a system cause the processor to perform method for unsupervised word image clustering comprising:

• capturing one or more image using at least one image capture device, wherein at least one of the one or more image comprises at least one word image;
• extracting one or more feature vector for each of the at least one word image using an untrained convolution neural network architecture, wherein extraction comprises:
• applying a first convolution to the at least one input image using a first plurality of filters, wherein the first plurality of filters are initialized by random filter based deep learning techniques using Gaussian random variable with zero mean and unit standard deviation, and wherein the plurality of filters are constrained to sum to zero,
• applying a Rectified Linear Unit (ReLU) non-linearity to a first plurality of feature maps, wherein the first plurality of feature maps is generated as output of the first convolution,
• applying a first sub-sampling to increase the position invariance of the first plurality of feature maps, wherein sub-sampling comprises applying a pooling operation on non overlapping segments of the first plurality of feature maps after application of the ReLU non-linearity,
• applying a second convolution on the output of the first sub-sampling using a second plurality of filters wherein the second plurality of filters are initialized by random filter based deep learning techniques using Gaussian random variable with zero mean and unit standard deviation, and wherein the plurality of filters are constrained to sum to zero,
• applying a Rectified Linear Unit (ReLU) non-linearity to a second plurality of feature maps, wherein the second plurality of feature maps are generated as output of the second convolution,
• applying a second sub-sampling to increase the position invariance of the second plurality of feature maps, wherein sub-sampling comprises applying a pooling operation on non-overlapping segments of the second plurality of feature maps after application of the ReLU non linearity, and
• combining the plurality of feature maps generated at the output of the second subsampling for extracting one or more feature vector;
• and
• clustering the one or more word images is based on the one or more feature vector, wherein the clustering comprises:
• clustering word images based on a similarity value across a first stage cluster, wherein the similarity value is evaluated by comparing the extracted one or more feature vector with a pre-defined first similarity threshold value (Tstrong);
• clustering word images based on a similarity value across a second stage cluster, wherein the similarity value is evaluated by comparing the extracted one or more feature vector, with a pre-defined second similarity threshold value (Tsoft);
• and
• wherein the first similarity threshold value (Tstrong) is higher than the second similarity threshold value (Tsoft),
• wherein the clustered word images are used for performing one or more functions, wherein the one or more functions comprises indexing, transcription, and optical character recognition.