Fakhri Karray
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fakhri Karray has a total of 180 co-authors. Here are the top co-authors by number of shared publications:
- Mark Crowley
Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON, Canada
55 shared publications - Benyamin Ghojogh
University of Waterloo
50 shared publications - Ali Ghodsi
University of Waterloo
31 shared publications - Milon Islam
Department of Computer Science and Engineering, Khulna University of Engineering and Technology, Khulna, Bangladesh
10 shared publications - Abdulmotaleb El Saddik
University of Ottawa
5 shared publications Benyamin Ghojogh
Department of Electrical and Computer Engineering Machine Learning Laboratory University of Waterloo Waterloo ON Canada
5 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
AdaptPrompt: Parameter-Efficient Adaptation of Vision-Language Models for Generalizable Deepfake Detection
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessAdaptPrompt: Parameter-Efficient Adaptation of Vision-Language Models for Generalizable Deepfake Detection
- Open Access
- 4 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 800-808
DatasetOS: An Agentic Framework For Multimodal Dataset Construction
- 4 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 800-808
Open AccessDatasetOS: An Agentic Framework For Multimodal Dataset Construction
- 13 September 2026
- Published by Springer Nature
- p. 60-77
SPQR: A Multi-dimensional Benchmark for Safety Alignment Under Benign Model Adaptation
- 13 September 2026
- Published by Springer Nature
- p. 60-77
SPQR: A Multi-dimensional Benchmark for Safety Alignment Under Benign Model Adaptation
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 3282-3293
SetLLM: Set Large Language Model for Cold-Start Item Recommendation
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 3282-3293
Open AccessSetLLM: Set Large Language Model for Cold-Start Item Recommendation
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Multimodal Deep Learning Fusion Approaches for Automated Depression Detection: A Systematic Review
Background: Depression affects over 280 million people worldwide, yet diagnosis still depends on subjective clinical interviews and self-report questionnaires that are vulnerable to recall bias, social desirability effects, and limited access to trained professionals. Multimodal deep learning may enable more objective depression detection by combining behavioral, verbal, and physiological signals from multiple data sources. Methods: This systematic review followed the PRISMA 2020 guidelines. We searched IEEE Xplore, ACM Digital Library, Web of Science, SpringerLink, and ScienceDirect for peer-reviewed studies published between January 2018 and August 2025. Eligible studies used at least two measured input modalities for depression classification or severity assessment with quantitative evaluation. From an initial pool of 120 records, screening and eligibility assessment yielded 50 included studies. Results: Across clinical, social media, and laboratory settings, multimodal systems outperformed unimodal baselines. Audio-video fusion and electroencephalography combined with facial video or eye tracking produced the largest gains. Alignmentaware audio-text models improved accuracy when reliable timestamps were available. Audio-video pipelines on standard corpora reached precision values near 95%, with sensitivity and specificity frequently above 80%, though metric reporting varied across studies. Accuracy generally plateaued beyond two or three modalities; well-optimized bimodal systems often matched or outperformed larger combinations while remaining easier to deploy. Among fusion strategies, hybrid approaches that combine feature-level and decision-level information most often achieved the highest accuracy and stability, and attention-based or transformer-based fusion added gains in both performance and interpretability. Conclusions: Multimodal deep learning can improve the objectivity of depression assessment, but clinical adoption requires progress on several fronts: standardized evaluation protocols and metrics, transparent handling of temporal alignment and missing modalities, external validation in diverse populations, fairness and uncertainty evaluation, and deployment-oriented research on privacy-preserving and on-device inference. This systematic review was not prospectively registered.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMultimodal Deep Learning Fusion Approaches for Automated Depression Detection: A Systematic Review
Background: Depression affects over 280 million people worldwide, yet diagnosis still depends on subjective clinical interviews and self-report questionnaires that are vulnerable to recall bias, social desirability effects, and limited access to trained professionals. Multimodal deep learning may enable more objective depression detection by combining behavioral, verbal, and physiological signals from multiple data sources. Methods: This systematic review followed the PRISMA 2020 guidelines. We searched IEEE Xplore, ACM Digital Library, Web of Science, SpringerLink, and ScienceDirect for peer-reviewed studies published between January 2018 and August 2025. Eligible studies used at least two measured input modalities for depression classification or severity assessment with quantitative evaluation. From an initial pool of 120 records, screening and eligibility assessment yielded 50 included studies. Results: Across clinical, social media, and laboratory settings, multimodal systems outperformed unimodal baselines. Audio-video fusion and electroencephalography combined with facial video or eye tracking produced the largest gains. Alignmentaware audio-text models improved accuracy when reliable timestamps were available. Audio-video pipelines on standard corpora reached precision values near 95%, with sensitivity and specificity frequently above 80%, though metric reporting varied across studies. Accuracy generally plateaued beyond two or three modalities; well-optimized bimodal systems often matched or outperformed larger combinations while remaining easier to deploy. Among fusion strategies, hybrid approaches that combine feature-level and decision-level information most often achieved the highest accuracy and stability, and attention-based or transformer-based fusion added gains in both performance and interpretability. Conclusions: Multimodal deep learning can improve the objectivity of depression assessment, but clinical adoption requires progress on several fronts: standardized evaluation protocols and metrics, transparent handling of temporal alignment and missing modalities, external validation in diverse populations, fairness and uncertainty evaluation, and deployment-oriented research on privacy-preserving and on-device inference. This systematic review was not prospectively registered.
- 29 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (2) , 3186-3198
Distilling Knowledge to Efficient Transformer for Semi-Supervised Citrus Maturity Detection Using Consumer UAVs
Accurate detection of citrus fruit maturity is critical for optimizing harvest schedules and maximizing yield. Consumer-grade unmanned aerial vehicles (UAVs) have emerged as cost-effective alternatives to traditional methods for detecting maturity, which rely on labor-intensive manual inspections. This paper presents a two-step, semi-supervised approach leveraging knowledge distillation (KD) and transfer learning for citrus maturity detection in UAV images. Specifically, we combine teacher-filtered pseudo-labels with a consistency-guided feature distillation signal to exploit abundant unlabeled UAV frames while using only a small labeled seed set. Firstly, a consistency-guided KD transfers knowledge from a pretrained detection transformer with collaborative hybrid assignment training (Co-DETR) to a lightweight student network by exploiting a small labeled and a large unlabeled dataset. The student network (Cit-DETR) is based on the highly efficient detection transformer (RT-DETR) having a ResNet18 backbone with selective kernel blocks and the hybrid encoder module. Step 2 uses a small labeled augmented dataset with maturity labels to fine-tune the Cit-DETR model for maturity detection. Experimental results on a custom UAV-captured citrus dataset demonstrate the effectiveness of our method, achieving 86.2% average precision in citrus detection and 91.0% mean average precision in ripeness detection. The model has been further optimized for real-time inference on edge devices or UAVs, enabling precision agriculture applications.
- 29 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (2) , 3186-3198
Distilling Knowledge to Efficient Transformer for Semi-Supervised Citrus Maturity Detection Using Consumer UAVs
Accurate detection of citrus fruit maturity is critical for optimizing harvest schedules and maximizing yield. Consumer-grade unmanned aerial vehicles (UAVs) have emerged as cost-effective alternatives to traditional methods for detecting maturity, which rely on labor-intensive manual inspections. This paper presents a two-step, semi-supervised approach leveraging knowledge distillation (KD) and transfer learning for citrus maturity detection in UAV images. Specifically, we combine teacher-filtered pseudo-labels with a consistency-guided feature distillation signal to exploit abundant unlabeled UAV frames while using only a small labeled seed set. Firstly, a consistency-guided KD transfers knowledge from a pretrained detection transformer with collaborative hybrid assignment training (Co-DETR) to a lightweight student network by exploiting a small labeled and a large unlabeled dataset. The student network (Cit-DETR) is based on the highly efficient detection transformer (RT-DETR) having a ResNet18 backbone with selective kernel blocks and the hybrid encoder module. Step 2 uses a small labeled augmented dataset with maturity labels to fine-tune the Cit-DETR model for maturity detection. Experimental results on a custom UAV-captured citrus dataset demonstrate the effectiveness of our method, achieving 86.2% average precision in citrus detection and 91.0% mean average precision in ripeness detection. The model has been further optimized for real-time inference on edge devices or UAVs, enabling precision agriculture applications.
- Open Access
- 15 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Computer Society
- Vol. 7, 365-375
AG-CLIP: Attribute-Guided CLIP for Zero-Shot Fine-Grained Recognition
Zero-shot fine-grained recognition is challenging due to high visual similarities between classes and the inferior encoding of fine-grained features in embedding models. In this work, we present an attribute-guided Contrastive Language-Image Pre-training (AG-CLIP) model with an additional attribute encoder. Our approach first identifies relevant visual attributes from the textual class descriptions using an attribute mining module leveraging a large language model (LLM) GPT-4o. The attributes are then used to construct prompts for an open vocabulary object/region detector to extract relevant corresponding image regions. The attribute text, along with focused regions of the input, then guides the CLIP model to focus on these discriminative attributes during fine-tuning through a context-attribute fusion module. Our attribute-guided attention mechanism allows CLIP to effectively disambiguate fine-grained classes by highlighting their distinctive attributes without requiring fine-tuning or additional training data on unseen classes. We evaluate our approach on the CUB-200-2011 and plant disease datasets, achieving 73.3% and 84.6% accuracy, respectively. Our method achieves state-of-the-art zero-shot performance, outperforming prior methods that rely on external knowledge bases or complex meta-learning strategies. The strong results demonstratethe effectiveness of injecting generic attribute awareness into powerful vision-language models like CLIP for tackling fine-grained recognition in a zero-shot manner.
- 15 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Computer Society
- Vol. 7, 365-375
Open AccessAG-CLIP: Attribute-Guided CLIP for Zero-Shot Fine-Grained Recognition
Zero-shot fine-grained recognition is challenging due to high visual similarities between classes and the inferior encoding of fine-grained features in embedding models. In this work, we present an attribute-guided Contrastive Language-Image Pre-training (AG-CLIP) model with an additional attribute encoder. Our approach first identifies relevant visual attributes from the textual class descriptions using an attribute mining module leveraging a large language model (LLM) GPT-4o. The attributes are then used to construct prompts for an open vocabulary object/region detector to extract relevant corresponding image regions. The attribute text, along with focused regions of the input, then guides the CLIP model to focus on these discriminative attributes during fine-tuning through a context-attribute fusion module. Our attribute-guided attention mechanism allows CLIP to effectively disambiguate fine-grained classes by highlighting their distinctive attributes without requiring fine-tuning or additional training data on unseen classes. We evaluate our approach on the CUB-200-2011 and plant disease datasets, achieving 73.3% and 84.6% accuracy, respectively. Our method achieves state-of-the-art zero-shot performance, outperforming prior methods that rely on external knowledge bases or complex meta-learning strategies. The strong results demonstratethe effectiveness of injecting generic attribute awareness into powerful vision-language models like CLIP for tackling fine-grained recognition in a zero-shot manner.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (9) , 1-14
Behavior-Aware Consistent Distillation for Cold-Start Recommendation
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (9) , 1-14
Behavior-Aware Consistent Distillation for Cold-Start Recommendation
- Open Access
AdaptPrompt: Parameter-Efficient Adaptation of VLMs for Generalizable Deepfake Detection
Recent advances in image generation have led to the widespread availability of highly realistic synthetic media, increasing the difficulty of reliable deepfake detection. A key challenge is generalization, as detectors trained on a narrow class of generators often fail when confronted with unseen models. In this work, we address the pressing need for generalizable detection by leveraging large vision-language models, specifically CLIP, to identify synthetic content across diverse generative techniques. First, we introduce Diff-Gen, a large-scale benchmark dataset comprising 100k diffusion-generated fakes that capture broad spectral artifacts unlike traditional GAN datasets. Models trained on Diff-Gen demonstrate stronger cross-domain generalization, particularly on previously unseen image generators. Second, we propose AdaptPrompt, a parameter-efficient transfer learning framework that jointly learns task-specific textual prompts and visual adapters while keeping the CLIP backbone frozen. We further show via layer ablation that pruning the final transformer block of the vision encoder enhances the retention of high-frequency generative artifacts, significantly boosting detection accuracy. Our evaluation spans 25 challenging test sets, covering synthetic content generated by GANs, diffusion models, and commercial tools, establishing a new state-of-the-art in both standard and cross-domain scenarios. We further demonstrate the framework's versatility through few-shot generalization (using as few as 320 images) and source attribution, enabling the precise identification of generator architectures in closed-set settings.
Open AccessAdaptPrompt: Parameter-Efficient Adaptation of VLMs for Generalizable Deepfake Detection
Recent advances in image generation have led to the widespread availability of highly realistic synthetic media, increasing the difficulty of reliable deepfake detection. A key challenge is generalization, as detectors trained on a narrow class of generators often fail when confronted with unseen models. In this work, we address the pressing need for generalizable detection by leveraging large vision-language models, specifically CLIP, to identify synthetic content across diverse generative techniques. First, we introduce Diff-Gen, a large-scale benchmark dataset comprising 100k diffusion-generated fakes that capture broad spectral artifacts unlike traditional GAN datasets. Models trained on Diff-Gen demonstrate stronger cross-domain generalization, particularly on previously unseen image generators. Second, we propose AdaptPrompt, a parameter-efficient transfer learning framework that jointly learns task-specific textual prompts and visual adapters while keeping the CLIP backbone frozen. We further show via layer ablation that pruning the final transformer block of the vision encoder enhances the retention of high-frequency generative artifacts, significantly boosting detection accuracy. Our evaluation spans 25 challenging test sets, covering synthetic content generated by GANs, diffusion models, and commercial tools, establishing a new state-of-the-art in both standard and cross-domain scenarios. We further demonstrate the framework's versatility through few-shot generalization (using as few as 320 images) and source attribution, enabling the precise identification of generator architectures in closed-set settings.