Source: https://daphnislabs.com/services/computer-vision

# Computer Vision Development

Image and video systems for detection, segmentation and visual inspection, integrated with the cameras and review tools your operation uses.

[Discuss your project](https://daphnislabs.com/contact?service=computer-vision) [Try the example](https://daphnislabs.com/services/computer-vision#worked-example)

Inspection camera / frame 017

|||| ||| || |||||

Label region

**Review image** Glare crosses the barcode

## Tools & technologies

-   OpenCV
-   Python
-   Vision models
-   Camera or upload APIs

## Make images useful to an operation.

### Object detection

Locate and count defined objects under specified camera conditions.

### Visual inspection

Compare images with labelled examples of acceptable and unusual states.

### Document capture

Extract image regions and text for downstream validation.

## Beyond a bounding box.

-   **Annotation specification**
-   **Vision inference pipeline**

-   **Condition-based evaluation report**
-   **Review interface and event integration**

## Check a package label

Capability Example

A camera image shows a parcel with a potentially damaged barcode label.

### Locate

### Read

### Escalate

Illustrative data. Nothing is sent to an external system.

Example workspace 1 / 3

### Captured frame

Illustrative spatial reference with a selected object

-   Camera: line-02
-   Object: parcel label
-   Glare detected across barcode

Detect the label region and check whether the image is usable.

### Barcode check

Illustrative spatial reference with a selected object

-   Label region cropped
-   Barcode partly unreadable
-   Shipment match: not confirmed

Decode the barcode and compare it with the expected shipment.

### Operator queue

Illustrative spatial reference with a selected object

-   Review required
-   Image and region attached
-   No automatic rejection

Send unreadable or mismatched labels to an operator.

A matched shipment ID or a review item carrying the source crop and reason.

Follow the record through the next step

## Start with the real camera conditions.

### Capture conditions

Evaluate lighting, movement, occlusion and camera placement using representative images.

### Retention

Define image storage, access and deletion rules before collecting production footage.

## Questions before we start.

### Can you use our existing cameras?

We first check resolution, frame access, placement and network constraints. Hardware replacement is a separate decision.

### How is accuracy assessed?

Use task-specific measures such as precision and recall, broken down by operating condition and the cost of missed or false detections.

[AI Development](https://daphnislabs.com/services/ai-development) [Deep Learning Development](https://daphnislabs.com/services/deep-learning) [TensorFlow Development](https://daphnislabs.com/services/tensorflow-development) [AI Integrations & APIs](https://daphnislabs.com/services/ai-integrations-apis)

## Bring your starting point.

Share representative images and the actions a correct or uncertain detection should trigger.

[Discuss your project](https://daphnislabs.com/contact?service=computer-vision) [Explore related services](https://daphnislabs.com/services/ai-development)