Without an account
- Use available detection models.
- History stays in this browser.
- No account photo archive or Telegram alerts.
CityCAM connects people who create visual detection models with people who use them. Choose a published model, point your phone camera and let it watch for the object or situation that matters to you.
CityCAM is not limited to cars. A model learns a specific visual target from example photographs: an object, a visible feature or a scene condition. What the app can detect depends on the models available in its catalogue, their training data and the camera view.
Examples of targets a creator could train include a tow truck in a street view, a particular vehicle attachment, an animal, an object at an entrance or a visible change at a workstation. These are possible applications, not a promise that every example already has a published model. Check the catalogue for active models, their purpose, supported markets and limitations.
Browse models by country and category. Read what each model is designed to detect.
Start the camera and detection together. The selected models analyse images on your phone; CityCAM does not continuously stream your video to a server.
See detections with a photo, time and model score. Premium adds Telegram alerts and a longer account history.
Creator Studio is a no-code tool: turn your own photographs into an AI detection model with a step-by-step wizard. No coding required.
You do not need to code, install TensorFlow or bring a ready-made AI model. Bring an idea and photographs you have permission to use. In the wizard, label examples with and without the target, start training and test the model on your phone. CityCAM handles the technical model setup. Creators choose varied, useful examples and check the results; uploading photos alone does not guarantee accuracy. Creation is available after creator approval, and publication requires model review.
The creator programme allocates 25% of eligible net Premium revenue to a shared monthly pool, divided by verified paid use of creators’ models. Rewards are subject to the programme terms; earnings are not guaranteed.
Creator programme terms (English)Build and train on a laptop or desktop, then open the same draft on a phone for field testing. Your work can be saved between steps.
Choose an existing category and a market, describe the target, add a preview and import photographs. Organise them into scenes with and without the target, including missed targets and false alarms. The wizard helps prepare separate training, validation and test scenes.
Train in your browser, then run field tests on a phone. Label captures, review the results and retrain a draft. The submission checklist shows the requirements still to meet before review.
When you enable a request for training contributions, users can voluntarily share photographs for that model. Sort them into target present, target absent or discarded, then use suitable examples in a new version. Sharing is separate from normal detection alerts.
Submit a version for administrator review. Published model weights stay fixed; improvements go into a new version. You can edit the public description and preview, share a direct link or QR code and review available link statistics. A shared link selects the model without starting the camera.
Your Studio brings together model states, usage, user star ratings, feedback, reward statements and payout details. Notifications highlight items needing attention. Approved creators can join a private Telegram community with topic previews in Studio.
The platform is under active development and testing. The standard creator workflow and user features described above are implemented; this does not certify any individual model’s accuracy. Purchase and payout availability depend on the account and the current service configuration.
A readiness checklist, AI pre-check and GPU optimisation workflow have been implemented for qualifying creator datasets. Processing is not generally enabled: activation and end-to-end testing are still pending. An optimisation produces a candidate version to compare and review, not an automatic accuracy guarantee. Synthetic image generation and independent recurring AI quality audits are not available yet.
No. CityCAM is a platform for creator-made visual detection models. Vehicles are one possible use case; the actual targets are defined by each model’s training data and description.
No coding or personal training server is needed for the standard wizard. The phone running CityCAM also runs detection. Premium alerts go to your linked Telegram recipient, which can be on another device; you do not need to pair two camera phones.
No. The displayed percentage is the model’s score for a frame. Accuracy must be evaluated on independent examples and real use. User star ratings express user feedback and are not a measurement of detection accuracy.
No. Useful, correctly labelled examples from varied independent scenes matter more than repeated similar images. Keep evaluation scenes separate from training, and test in the lighting, distance and camera view in which the model will be used.
The shared monthly pool receives 25% of eligible net Premium revenue. Your share depends on verified paid model use, not link clicks, registrations, star ratings or the number of models you publish. Payouts follow the programme terms and eligibility checks; income is not guaranteed.
You can use CityCAM anywhere in the world. The platform connects creators and users across countries, with models organised by market and category. We welcome creators who understand local needs and bring useful, inventive and enjoyable models to their communities. Availability of individual models and features may vary by market.
Terms of Service (English)Camera permission, an open app and an unlocked phone are needed. Internet is required for account photo uploads and Telegram delivery.
Lighting, distance, viewpoint and training examples affect results. A model’s percentage is its score, not a guarantee of accuracy.
Some creators request photographs to improve their models. Sharing requires separate, voluntary consent for the model, and can be withdrawn. It is not enabled just by selecting a model.
The read-only REST API lets Premium accounts retrieve their own retained detection events, authorized photographs and available models through CityCAM’s servers. Account API keys have selectable permissions, expiry and immediate revocation; a synthetic connection test also works without Premium. Signed webhooks can deliver new confirmed detections to verified HTTPS endpoints, with bounded retries and delivery status. The API also reports open app instances and supports scoped model-selection and detection-restart commands with expiry and device acknowledgement. It cannot start a stopped camera, wake a closed app or reboot a phone. Potential uses include operational or industrial monitoring; CityCAM is not a certified safety-critical industrial sensor.