Object image retrieval based on self-supervised learning and zero-shot learning

SeCo-SBIR: Semantically Consistent Prompt Learning for Zero-Shot Sketch-Based Image Retrieval

SeCo-SBIR is presented, a semantically consistent prompt learning framework that achieves state-of-the-art results on all three standard ZS-SBIR benchmarks across categorical, generalized, and across-dataset settings.

Attention-enhanced vision transformer hashing for hybrid image retrieval

Large-scale image retrieval requires compact representations without substantially sacrificing retrieval accuracy. However, Vision Transformer Hashing (VTS) concatenates all output tokens before hash projection, resulting in a high-dimensional hashing head with considerable model and memory overhead. We replace this token-concatenation operation with learnable-query multi-head attention pooling, which aggregates ViT tokens into a compact, image-adaptive representation for supervised hash-code learning. We further evaluate the proposed model using a standard two-stage retrieval procedure. In Stage 1, the attention-pooled representation is mapped to a binary code for efficient Hamming-space candidate selection. In Stage 2, the final-layer CLS descriptor from the same shared ViT-B/16 is used to re-rank the shortlisted candidates by cosine similarity. Controlled experiments on MS-COCO and NUS-WIDE compare attention pooling with GeM, average, CLS, and VTS-style concatenation. The proposed hybrid configuration achieves 91.13% and 88.63% mAP@5000 on MS-COCO and NUS-WIDE, respectively, demonstrating performance comparable to the VTS-style Concatenation baseline under the same controlled configuration, with marginal gains of 0.15 and 0.31 percentage points. It also exceeds separately trained 32-bit attention hash-only models by 2.23 and 1.30 points. Relative to concatenation, attention pooling reduces model footprint and batch-1 GPU memory by 63.1% and 62.0%, respectively, while increasing throughput by 21.7%. These results show that attention pooling substantially improves the efficiency of VTS while preserving retrieval effectiveness, whereas continuous re-ranking provides most of the gain over Hamming-only retrieval.

G2D: Generative-to-Discriminative Collaborative Inference for Zero-Shot Image Classification

G2D is proposed, a training-free framework that uses a generative VLM to verify CLIP-retrieved candidates against the image and transfers to DCLIP, WaffleCLIP, and CuPL, supporting a practical interface between discriminative proposal and generative visual reasoning.

Few-shot image classification algorithm based on deep learning and feature fusion

Few-shot image classification remains difficult because a model must identify novel classes from only one or a few labeled examples while preserving discriminative local information. Metric-learning methods based on Earth Mover’s Distance (EMD) improve local correspondence by representing an image as a set of regional embeddings, but standard backbones treat feature channels uniformly and may retain irrelevant responses. This study proposes ECA-DeepEMD, a channel-attentive metric-learning framework that inserts efficient channel attention (ECA) into a ResNet-12 feature extractor and then computes structural similarity through differentiable EMD. The ECA module captures local cross-channel dependencies without dimensionality reduction, allowing the embedding network to emphasize semantically informative channels with limited computational overhead. The model was evaluated under 5-way 1-shot and 5-way 5-shot protocols on Mini-ImageNet, Tiered-ImageNet, and FC100. It achieved accuracies of 67.14% and 84.90% on Mini-ImageNet, 73.51% and 88.54% on Tiered-ImageNet, and 48.20% and 65.58% on FC100, respectively. Against the reported DeepEMD baseline, the gains on Mini-ImageNet were 1.23 and 2.49 percentage points. Grad-CAM visualizations further indicate that channel attention concentrates responses on category-relevant regions. These findings show that lightweight channel recalibration complements optimal-transport-based local matching and provides a practical approach to data-constrained visual recognition.

Rethinking Text-Based Image Retrieval in Specific Domain

The Semantic-Aware Fine-Tuning (SAFT) framework is proposed to address semantic compression in specific domains, which incorporates Semantic-Aware Soft-Label Supervision and Intra-modal Structural Distillation to establish a promising paradigm for domain-specific TBIR tasks.

Hybrid CNN-ViT feature embedding for image retrieval with multi-loss optimization