This is a machine learning based desktop app that was built for Indian Railways. The app is used to track the spark generated in train pantograph. This helps to detect undulation in track and overhead equipments. The app uses a custom built YOLOv11 model, advance sensors, external cameras to detect the spark and train movement.

Accurate tracking of the spark and its calculation was paramount. We used cutting edge algorithm to detect spark in varying situations in day and night.

The app supported multitude of sensors : GPS for location, UV for light intensity, Tachometer for accurate speed. The app supports external cameras and provides advance customization for the webcam.

Below are the main highlights :

• Track/Route Data : Ability to accept train route data and map them to track the train's location, direction and plot it on map.
• Sensors : Cameras, GPS sensor, UV Sensor, Tachometer. The app was modular and can support multiple sensors and peripherals. The data from the sensors was used to give live analytics on train Information. As well as it was used to create detailed report.
• Machine vision : We trained custom model with thousands of images. Our model was super accurate and could detect more than 99% of sparks. But we wanted to decrease gpu usage as much as possible so we also employed plethora of advance algorithm to detect spark with CPU. In challenging conditions the app switched to the GPU.
• Video/GIF/Images : Every frame was labeled with the data coming from multiple sensors. We generated GIFs, Images and Continuous Videos for the sparks. This helped to audit the train route.
• Detailed report : User can get detailed report with ability to customize report layout and filter report data. The report also contained the labeled images. The report can be downloaded in PDF as well as Excel formats.
• License Management : We created a rust desktop app that can generate secure license keys to safeguard and license the software.