####Read it Until the End My current Python-based PaddleOCR system works with perfect accuracy in CPU mode. However, each operation takes around 1.5–2 minutes. When switched to GPU mode, the process time drops to approximately 40 seconds, but in rare cases, the order of odds headers becomes misaligned. --- ### Required Modifications 1. GPU Mode Optimization The GPU mode should work with the same accuracy as the CPU mode. * The order of the odds headers must remain consistent. * The only difference should be the speed improvement. 2. Integration of 3 New Fields from a Different Page Three additional data fields from a different tab within the mobile app must be extracted using OCR and properly added to the output panel. * Detailed specifications will be provided privately. * The page may require scrolling; appropriate logic should be implemented. --- ### Environment and Requirements The full project folder with all necessary tools installed in a virtual environment will be provided. * Python version: 3.10 * Pip version: 25.2 * PaddleOCR version: 2.7 * GPU mode requires a proper setup with CUDA and cuDNN. It will not work without them. You may use Samsung Flow, BlueStacks, or LDPlayer to access and interact with the mobile app. --- ### Important Notes * The mobile application must be in Turkish. * A VPN located in Turkey is required to access specific content within the app. --- ### Please Note If you intend to change your offer after applying, do not apply. Submit your final and fixed offer up front. I am not interested in wasting time with negotiations after submission. My budget is limited, so I kindly ask you to submit reasonable offers. Delivery time: within 1–2 days.