Alfonso Blanco.

Recognition of license plate numbers, in any format, by automatic detection with Yolov8, pipeline of filters and paddleocr as OCR

From dataset https://universe.roboflow.com/drone-detection-pexej/drone-detection-data-set-yolov7/dataset/1 a model is obtained, based on yolov10 to detect drones in images. Predictions from several models are used in cascade to obtain the optimal result.

Drone detection with a simple program obtained by asking a simple question in the address bar of bings (copilot search)

Since have been noticed that Roboflow's RFDETRBase incorporates the drone class as an additional class, this drone detection program is presented.

Classification of skin lesions (among 7 classes) using the file https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/DBW86T and using the pytorch resnet model. The success rate for the specific test file (unseen data) that comes with the download file is 81.13%.

From dataset https://universe.roboflow.com/test-svk7h/brain-tumors-detection/dataset/2 a model is obtained, based on yolov10 to indicate tumors in images of brains.

Brain tumor detector using the Roboflow dataset https://universe.roboflow.com/test-svk7h/brain-tumors-detection/dataset/2 as a training dataset and simple programs. It requires very few resources, and training can be run on a personal computer.

Brain tumor detector using the Roboflow dataset https://universe.roboflow.com/test-svk7h/brain-tumors-detection/dataset/2 as training dataset and simple programs generated by querying Bing's AI. It requires very few resources and can be run on a personal computer.

Car license plate recognition test using fast_plate_ocr

A test to detect car license plates using Ollama.

Test for applying object detection in maritime aerial views with RFDETRBase, using https://universe.roboflow.com/jacob-solawetz/aerial-maritime/dataset/24 ​​as the custom dataset.

solar_panel_anomalies-Yolo. This is an essay that obtains a model to detect anomalies in solar panels using the roboflow file https://universe.roboflow.com/ron-zhyan/solar-panel-anomalies-hikbk-0joqn/dataset/1 as a dataset and a training with yolov11

Project to detect errors in label printing using an autoencoder. Use the Roboflow file as both the training and test files: https://universe.roboflow.com/university-science-malaysia/label-printing-defect-version-2/dataset/25

This work is an extension of the project https://github.com/ablanco1950/LicensePlate_Yolov8_Filters_PaddleOCR adding the possibility to detect the speed, tracking and counting cars. Has been updated to feature and leverage improvements brought by Roboflow and Ultralytics.

This is an experiment adapting the project https://github.com/hubert10/fasterrcnn_resnet50_fpn_v2_new_dataset to detect wrist fractures using a selection of wrist fracture images obtained from https://universe.roboflow.com/landy-aw2jb/fracture-ov5p1/dataset/1 as a custom dataset. Given the positive results, it is uploaded to GitHub.

Test using a simplified Yolo model from Scratch, training the Roboflow file https://universe.roboflow.com/drone-detection-pexej/drone-detection-data-set-yolov7/dataset/1 and using OpenCV functions for drone detection.

Detection of fractures in images by obtaining the X and Y coordinates of the center of the fracture applying ML (SVR). It is applied to a selection of data from the Roboflow file https://universe.roboflow.com/landy-aw2jb/fracture-ov5p1/dataset/1 Compared to other tests using DL for the same set of data, much better precision and training time

LicensePlate_Yolov8_MaxFilters: recognition of car license plates that are detected by Yolov8 and recognized with pytesseract after processing with a pipeline of filters choosing the most repeated car license plate. In a test with 21 images, 18 hits are achieved

Project that estimates the distance a car is on a road based on the relationship between the real size of the car and the size it appears in the video obtained. It also estimates the lane the car are traveling in at any given time based on the angle between the position of the car and camera, even guess lane change intentions. Also included is a

Program that allows to determine the label of a sonar signal, whether it corresponds to a mine or a rock, based on CV2 KMEANS. It starts from the signal file sonar_data.csv, although it is a labeled file, the labels are only used to check the clustering results. A success rate of 96.63% is obtained

It's a Wpod-net demo, downloaded from https://github.com/quangnhat185/Plate_detect_and_recognize, for the recognition of car license plates, the use of labeled images is avoided, with lower accuracy

Essay to detect targets in images obtained by radar. Use the images provided by https://github.com/kz258852/dataset_M_Radar and use yolov11 to create a model for detecting true targets in the images

Test that, based on the test file of the dataset https://github.com/moodoki/radical_sdk?tab=readme-ov-file, attempts to detect targets with an indication of their distance and speed.

From dataset https://universe.roboflow.com/roboflow-100/bone-fracture-7fylg a model is obtained, based on yolov10, with that custom dataset, to indicate fractures in x-rays. The project uses 5 cascade models, if one does not detect fracture it is passed to another

Project that uses Yolov8 as license plate detector, followed by a filter that is got selecting from a filters collection with a code assigned to each filter and predicting what filter with a CNN process

Config files for my GitHub profile.

# SUSY_WEIGHTED: from the SUSY.csv training file (https://archive.ics.uci.edu/ml/datasets/SUSY), create a weighted SUSY that, together with a program included in the project, allows obtain success rates higher than 90%. This precision can be increased by following the iterative method of the project.

kNN-MIN: A Spark-based design of the k-Neighbors Neighbors classifier for big data, using minimal resources and minimal code.

HASTIE_NAIVEBAYES: from the Hastie_10_2.csv file obtained by the procedure described in https://scikit-learn.org/stable/modules/generated/sklearn.datasets.make_hastie_10_2.html, obtains a success rate in the training of 88% and 84% in the test. The main difference is that in the statistical process, each field is sampled differently according to its contribution to the hit rate.

# SUSY_WEIGHTED_V1: from the SUSY.csv test file (https://archive.ics.uci.edu/ml/datasets/SUSY), create a valued SUSY.csv that, together with a program included in the project, allows obtain success rates higher than 78%. The main difference id that In the statistical process, each field is sampled differently according to its contribution to the hit rate.