AUS
NZ
UK
CA
USA
FRA
GER
ESP
ITA
NLD
PRT
SWE
GRC
IND
POL
BRA
ZAF
RUS
CHN
UAE
SAN JOSE, Calif., April 12, 2024 /PRNewswire/ -- INFINIQ, a leading South Korean AI platform services company, announced a breakthrough in few-shot object detection with the publication of their research paper, "Re-scoring using Image-Language Similarity for Few-Shot Object Detection," in the esteemed international journal, "Computer Vision and Image Understanding."
INFINIQ's RISF model outperforms conventional methods in object detection with minimal data
The paper explores the development of a novel AI model named RISF (Re-scoring using Image-Language Similarity for Few-Shot Object Detection). RISF addresses the challenge of accurately detecting objects in datasets containing limited images (less than 30). The model leverages the power of image-language similarity to pinpoint object location and classification.
RISF combines an object detection model with a Contrastive Language-Image Pre-training (CLIP) model. To ensure seamless integration and enhance accuracy, INFINIQ researchers developed a new loss function called Background Negative Re-scale Loss (BNRL).
INFINIQ's RISF achieved recognition by securing the second position in the few-shot object detection category on "papers with code," a renowned platform for sharing AI research, achieving an average precision (AP) score of 25.5. The paper's publication in Computer Vision and Image Understanding, a high-caliber SCI(E)-indexed journal, further validates the model's significance.
"RISF outperforms conventional methods in object detection with minimal data," said Min Jae Jung, lead researcher behind the paper. "This exceptional performance and accuracy make RISF a valuable model during the AI learning phase."
"The publication of RISF in such a well-respected journal signifies its global potential," commented Jun Hyung Park, CEO of INFINIQ. "INFINIQ remains committed to driving innovation in the field of artificial intelligence."
About INFINIQ
INFINIQ, a member company of Global Digital Innovation Network (formerly known as Born2Global Centre), is a specialized artificial intelligence platform company with expertise in autonomous driving development and computer vision fields, serving a diverse clientele. With a focus on the AI development platform 'AI Studio' and the AI-based video analysis solution 'Heidi AI', the company is expanding its footprint into the autonomous driving, security, and defense sectors. Moreover, it has been diversifying its offerings since its inception in 2016, starting with AI data processing services, and now encompassing various services based on AI Vision and AI platforms.
