A children's learning system that uses a camera to recognize animals with YOLO11, colors with HSV, and numbers with OCR, then shows and speaks the result in Arabic.
An academic project carried out by a team of students with technical assistance from Techno Enjaz to build an interactive learning system for children that uses a computer camera to recognize animals, colors, and numbers, then shows the result on screen and plays an Arabic audio file that pronounces it. The system relies on YOLO11 for animal detection, the HSV color space for color identification, and OCR for reading numbers, within a graphical interface built with Tkinter and Python.
The project documents detailed technical results for each function: the animal detection models recorded mAP50 values between 0.856 and 0.992 depending on the class, the system read 7 of 10 numbers correctly in a direct test, and color reading remains dependent on lighting and camera conditions. The project includes no educational study measuring its effect on children's learning.
| Item | Details |
|---|---|
| Project type | Interactive learning system / prototype |
| Field | Artificial intelligence, computer vision, early-learning technology |
| Project status | Academic prototype implemented and tested on a laptop |
| Techno Enjaz role | Technical support and assistance to the students during project development |
| Core technologies | Python, OpenCV, NumPy, YOLO11, OCR (EasyOCR), HSV, Tkinter, Playsound |
| Input | The computer's built-in camera |
| Outputs | The name of the animal, color, or number on screen, with pre-recorded Arabic audio |
The system is built on live interaction: the child shows an animal, a colored object, or a number to the camera, and the program recognizes it automatically and says its name clearly. This direct link between image and sound aims to strengthen the child's visual and language understanding in a way close to the sensory learning typical of early childhood.
The system was designed to be simple enough for a child to use without constant supervision, with Arabic audio content to develop pronunciation, and an interface suited to children in its colors, element sizes, and ease of interaction.
The project was carried out by the students within an academic framework, with technical assistance from Techno Enjaz during the development of the prototype and the testing of its components. The office's contribution included supporting the technical work and guiding the development process to help turn the idea into a testable prototype, while the academic execution of the project remained the students' responsibility.
The project reflects practical experience with computer vision applications and with connecting AI models to user interfaces and testable image-processing pipelines.
| Component | Role |
|---|---|
| Laptop | Built-in camera for live capture and internal speaker for audio output, with no extra devices |
| Python | Language used to develop the application and connect its components |
| OpenCV | Capturing and processing frames, converting them to HSV, and enhancing images before analysis |
| NumPy | Numerical operations on image arrays |
| YOLO11 | Detecting and classifying animals in the image |
| OCR (EasyOCR) | Extracting the number shown to the camera as text |
| Tkinter | The graphical interface and navigation between activities |
| Playsound | Automatically playing the pre-recorded audio file matching the result |
Each activity follows a clear processing path:
YOLO11 models were trained on custom animal-detection data, and each model was evaluated separately with Precision, Recall, mAP, and F1, along with the speed of each processing stage.
| Model | mAP50 | mAP50-95 | Best F1 (at confidence threshold) | Inference time |
|---|---|---|---|---|
| Eagle | 0.956 | 0.824 | 0.92 (0.409) | 18.8 ms |
| Cats | 0.992 | 0.784 | 0.98 (0.591) | 28.4 ms |
| Fish | 0.856 | 0.551 | 0.81 (0.449) | 21.2 ms |
These values reflect the detection models' performance within the project's tests and do not represent a single accuracy figure for the whole learning system. Inference time also depends on the hardware used in testing.
The color function reads the color value at a point in the center of the screen, converts it to the HSV space, and identifies the color name by comparing it with set thresholds for each color; the name is then shown and its audio played.
The function was tested manually by showing ten colors to the camera (red, green, blue, yellow, orange, brown, white, black, purple, and pink) and comparing the output name with the actual color. The report gives no final accuracy figure for this test; it includes only an illustrative example of how accuracy is calculated. The project identified the factors that affect color reading: lighting, camera type, shadows and reflections, similarity between some colors such as orange and brown, and the distance between the object and the camera.
The numbers function uses OCR (optical character recognition) to extract the number shown to the camera and convert it to text, then displays the number and plays its audio.
| Number shown | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| OCR reading | 1 | 2 | 3 | 4 | S | 6 | 7 | B | 9 | 1O |
In this direct test, 7 of 10 samples were read correctly (70%). Errors occurred on visually similar shapes: 5 was read as the letter S, 8 as B, and 10 was read with the letter O instead of zero.
The report also presents a separate evaluation of EasyOCR performance on number detection: Accuracy of 87.4%, Precision of 0.89, Recall of 0.83, and F1 of 0.86, with 45 false positives and 81 false negatives. Because this evaluation's context differs from the ten-digit test, the two results are presented separately and not merged into a single accuracy figure.
The project proposed ways to improve reading: even lighting without shadows, higher contrast between the number and the background, clear fonts such as Arial or OCR-B, larger numbers in the frame, a straight camera angle, and programmatic correction of common errors such as converting O to 0 and B to 8.
The project identified several directions for later versions:
These items are future proposals, not functions of the current version.
The system uses a camera pointed at the child and their surroundings, so in any real use it is preferable to keep processing local to the device and not to store children's images except with parental consent and under clear rules.
If you are working on an interactive computer-vision learning app and need technical help developing the idea or building a prototype, you can contact the Techno Enjaz office to discuss your project's requirements and the appropriate scope of support.
The system offers three functions: recognizing animals, colors, and numbers through the computer's camera. A single Tkinter interface ties them together; the system shows the result on screen and then plays a pre-recorded Arabic audio file that pronounces it. It was built by a team of students with technical assistance from Techno Enjaz.
The camera captures a frame that is preprocessed with OpenCV, then passed to a YOLO11 model for animal detection, converted to HSV and compared with per-color thresholds, or read with OCR to extract the number. The item's name is then displayed and the Playsound library plays the matching audio file automatically.
The eagle model recorded mAP50 of 0.956 and F1 of 0.92, the cats model mAP50 of 0.992 and F1 of 0.98, and the fish model mAP50 of 0.856 and F1 of 0.81, with inference times of 18.8, 28.4, and 21.2 ms respectively. These values reflect the detection models' performance in the project's tests, not a single accuracy figure for the whole system.
In a direct test of the digits 1 to 10, 7 samples were read correctly, with errors such as reading 5 as S and 8 as B, and the report also presents a separate EasyOCR evaluation at 87.4% accuracy. Colors were tested manually with ten colors without a final accuracy figure in the report, and they are affected by lighting, shadows, and camera type.
No. The available materials include no educational study measuring learning gains or the system's effect on children's learning, and results are limited to technical performance and documented software experiments. Additions such as speech recognition, multi-language support, adaptive learning, and augmented reality remain future proposals.
Reach out to Techno Enjaz engineering team to discuss technical feasibility, prototype hardware, and development scope.
Discover more engineering prototypes and academic implementations
A computer vision system that detects knives and pistols in real time from a laptop or phone camera using two trained YOLOv8 models, Python, and OpenCV.
An interactive learning system for children that recognizes objects via the camera with YOLOv8, speaks their names, detects colors with HSV, and labels uploaded images.
A gold-shop security system detecting weapons with YOLO, violence with an RNN and theft with MediaPipe, plus PIR and flame sensors on NodeMCU and Telegram alerts.
