Computer Vision is one area of artificial intelligence that has received unprecedented interest in the past decade owing to the vast scope of its applications. The ability of machines to acquire and process visual data has enabled very sophisticated solutions to practical problems faced in various industries. Consequently, there are a number of patents that have been filed to secure the exclusive rights to certain powerful uses of CV. Here, we list some of the most notable patent filings in computer vision.

• Recognition of data in one or more images (WIPRO, 2016): This method enables more effective decision making by enabling accurate text/object identification from images. The method receives the image to identify segmented objects. Further, it generates an object relationship data for each of the segmented objects and determines a knowledge base representation based on defined features. A Recurrent Neural Network (RNN) is trained based on the knowledge base representation to determine an appropriate Neural Network (NN) having the optimum confidence score. This is how objects in the input image are predicted and transmitted to external systems for decision making. The inventors include: Govindaraj Balaji, Mohd Zaid, Sujatha J, Raghottam Mannopantar.
• Extracting product information from low-resolution photos for updating product master (Infilect Technologies Private, 2019): This system is used for updating a global product-master database by extracting product information of assets from the low-resolution media contents. An image capturing device captures the media contents of assets. The information extraction system works as follows: (i) receives the media contents from the device, (ii) identifies the low-resolution media contents based on their size and dimensions, (iii) parses the low-res media contents using a system of cascading deep neural networks, to generate the product information at Stock Keeping Unit (SKU) level, (iv) transfers the information obtained to the point-of-sale (POS) device to update a local POS product-master, and (v) transmits the product information to the a global product-master server to update the global database. The inventors include: Gabale Vijay, Subramanian Anand Prabhu, Kannan Vijaykumar.
• Data acquisition and asset inspection in presence of magnetic interference (TCS, 2020): Data acquisition and assets inspection systems in many infrastructures such as railway, power line, and bridges provide inaccurate results in presence of magnetic interference. This system proposes a UAV based navigation through a dynamic correction path to inspect one or more assets in one or more infrastructures. A plurality of sensors are integrated with the UAV to acquire images in presence of the magnetic field. The acquired images are further processed to segment and detect anomalies in various parts of the asset. The detected anomalies are further classified as potential anomalies and non-potential anomalies. The proposed method provides accurate results with reduced processing time. The inventors include: Joshi Sunil Dattatraya, Mishra Mayank, Vyawahare Vaibhav, Salsingikar Shripad, Gubbi Lakshminarasimha Jayavardhana Rama, Kotamraju Srinivas, Bhogineni Sreehari Kumar.
• Generating a semantic representation of a target image for an image processing operation (QUALCOMM): Various embodiments may include a computing device analysing an image to identify various elements of interest in the image, identifying concepts associated with these elements of interest, and identifying potential elements of interest and potential concepts that are not included in the image using other information. This is done by generating the semantic representation of the target image based on the selected elements of interest and concepts. The inventors include: Krishnamurthi Govindarajan, Raman Arun

Join Neerav Parekh, Anand Subhash and Gagandeep Reehal in a power packed fireside chat during the AI Patent Power Hour, where they discuss India's potential & challenges involved to become a leader in filing AI patents.

Register here: