MétaCan
Menu
Back to cohort
Record W4406254008 · doi:10.31436/iiumej.v26i1.3268

4D Radar Imaging and Camera Fusion for Road Crossing Detection and Classification Using Deep Learning

2025· article· en· W4406254008 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

VenueIIUM Engineering Journal · 2025
Typearticle
Languageen
FieldComputer Science
TopicAdvanced Neural Network Applications
Canadian institutionsUniversity of Waterloo
Fundersnot available
KeywordsArtificial intelligenceComputer scienceComputer visionRadarDeep learningObject detectionArtificial neural networkPattern recognition (psychology)Telecommunications

Abstract

fetched live from OpenAlex

This paper presents the development of an object detection and classification system for road crossing areas, integrating 4D radar imaging and a mono-camera dataset with a deep-learning neural network. The system utilizes deep neural networks implemented via Keras and TensorFlow to detect and classify multiple targets, including pedestrians, cars, buses, and trucks. At the core of this work is Retina-4F, a multi-chip radar imaging system developed by Smart Radar System, which offers high-resolution object detection and localization capabilities. Retina-4F provides real-time 4D information on detected objects, operating in a cascading architecture with three transmitters and four receivers per chip. Two road-crossing scenes were simulated to collect data, generating a point cloud dataset labeled with target classes for neural network training and testing. Data from two main sensors—Retina-4F and a mono-camera—were pre-processed using DBSCAN and YOLOv7 for enhanced accuracy. Operating at 77 GHz, Retina-4F was tested in two road environments, generating a dataset with approximately 10,000 frames. The deep learning model demonstrated an accuracy of 84% in classifying multiple targets, including cars, pedestrians, buses, and trucks. The fusion of radar point cloud data with visual sensor data proved effective, showing strong results in distinguishing target types. ABSTRAK: Kertas ini membentangkan pembangunan sistem pengesanan dan pengelasan objek untuk kawasan lintasan jalan raya, menggabungkan pengimejan radar 4D dan set data mono-kamera dengan rangkaian neural pembelajaran mendalam. Sistem ini menggunakan rangkaian neural mendalam yang dilaksanakan melalui Keras dan TensorFlow untuk mengesan dan mengelaskan pelbagai sasaran, termasuk pejalan kaki, kereta, bas, dan trak. Inti daripada kajian ini adalah Retina-4F, sistem pengimejan radar berbilang cip yang dibangunkan oleh Smart Radar System, yang menawarkan keupayaan pengesanan objek dan penentuan lokasi resolusi tinggi. Retina-4F menyediakan maklumat 4D masa nyata mengenai objek yang dikesan, beroperasi dengan tiga pemancar dan empat penerima bagi setiap cip dalam seni bina kaskad. Dua adegan lintasan jalan disimulasikan untuk mengumpul data, menghasilkan set data awan titik yang dilabel dengan kelas sasaran untuk latihan dan ujian rangkaian neural. Data daripada dua sensor utama—Retina-4F dan mono-kamera—dipra-proses menggunakan DBSCAN dan YOLOv7 untuk meningkatkan ketepatan. Beroperasi pada 77 GHz, Retina-4F diuji dalam dua persekitaran jalan yang berbeza, menghasilkan set data dengan kira-kira 10,000 bingkai. Model pembelajaran mendalam menunjukkan ketepatan sebanyak 84% dalam mengelaskan pelbagai sasaran, termasuk kereta, pejalan kaki, bas, dan trak. Penggabungan data awan titik radar dengan data sensor visual terbukti berkesan, menunjukkan hasil yang kuat dalam membezakan antara jenis sasaran.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: none
Teacher disagreement score0.883
Threshold uncertainty score0.702

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.000
Science and technology studies0.0010.000
Scholarly communication0.0010.001
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.011
GPT teacher head0.259
Teacher spread0.248 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it