Launch HN: MoonVision https://ift.tt/2CQqa2p

Launch HN: MoonVision Dear HN, this is Alex, cofounder of Moonvision, an Austrian based computer vision company. We started in 2017 tracking grilled chicken at the Oktoberfest Munich [1] and transitioned into automating visual inspections tasks. Our web tools are used by quality assurance experts to manage training data and create custom models without an external workforce. For such tasks and experts, the effort to label data is often prohibitive. Therefore, we built tools that work with low amounts of initial data. To train a new pipeline we cover the following 6 steps: - Video gathering - Object mining - Snapshot detection - Annotation - Training - Deployment Training data Since video is the cheapest form of input, we extract relevant/repetitive scenes from it first. These scenes are cut to snapshots that avoid motion blur, occlusion and person specific information. Before an expert starts labelling, we create so-called auto entities. To further speed up annotation, we cluster entities interactively. Also, there are ways to check the data and isolate outliers. We use a Redis based engine for label recommendations and few-shot detection on HTTP endpoints. Our training pipeline can also jumpstart these processes by unsupervised training. Deployment Our product is currently used by many big german manufacturing companies like Audi. This requires CD of x86 and ARM images that work on- and offline. Currently, we are looking to partner with integrators of machine vision solutions abroad. With a strong background in applied research, we’ll make sure to deliver on the quintessential outcome: tools that take human-machine collaboration to the next level. You can find a illustrative examples on our website. As a computer vision practitioner we would like your feedback on the annotation tool. There is a free version to use for your own project [2] Danke [1] https://www.youtube.com/watch?v=iHUnsdzFrWQ [2] https://ift.tt/2Ua6Mbs April 2, 2019 at 07:20AM