Researchers in several disciplines and fields agree that the image establishing has both perceptual and affec-tive evaluations.The paper comparatively analyzes the image of landscapes in ancient water towns held by 16...Researchers in several disciplines and fields agree that the image establishing has both perceptual and affec-tive evaluations.The paper comparatively analyzes the image of landscapes in ancient water towns held by 1619 tourists after the investigations of Zhouzhuang and Tongli of Jiangsu Province.Based on the elements of the image of urban landscapes,the perceptual/cognitive image of tourist landscapes in water towns is developed including path,node,landmark,important courtyard,district and edge.Particularly the important courtyard plays the great role on the per-ceptive image of tourists,however,the perceptual/cognitive images of the district and the edge are obviously weak.Meanwhile,the finding showed that the affective images of tourist landscapes in Zhouzhuang differ with those of Tongli by the method of paired comparison.The main reason is that the affective images are influenced by the impressive scenery spots.Furthermore,the commercialization of streets and pollution of rivers are easy to be functioned negatively.展开更多
Image captioning models typically operate with a fixed vocabulary,but captioning is an open-vocabulary problem.Existing work addresses the image captioning of out-of-vocabulary words by labeling it as unknown in a dic...Image captioning models typically operate with a fixed vocabulary,but captioning is an open-vocabulary problem.Existing work addresses the image captioning of out-of-vocabulary words by labeling it as unknown in a dictionary.In addition,recurrent neural network(RNN)and its variants used in the caption task have become a bottleneck for their generation quality and training time cost.To address these 2 essential problems,a simpler but more effective approach is proposed for generating open-vocabulary caption,long short-term memory(LSTM)unit is replaced with transformer as decoder for better caption quality and less training time.The effectiveness of different word segmentation vocabulary and generation improvement of transformer over LSTM is discussed and it is proved that the improved models achieve state-of-the-art performance for the MSCOCO2014 image captioning tasks over a back-off dictionary baseline model.展开更多
Breast pathology is varied, bringing together tumor and non-tumor lesions. Objective: To study the contribution of the ultrasound-mammography pair in the diagnosis of breast pathologies. Materials and Method: This was...Breast pathology is varied, bringing together tumor and non-tumor lesions. Objective: To study the contribution of the ultrasound-mammography pair in the diagnosis of breast pathologies. Materials and Method: This was a retrospective descriptive study, carried out over a period of 3 years (from January 2018 to December 2020) at the Diagnostic Imaging Center (C.I.D) “TERIYA” in BAMAKO. It concerned all patients who came for a mammogram/ultrasound examination of the breast. All women admitted for mammogram or breast ultrasound who were diagnosed with a breast injury during the study period were included. Incomplete records and radiological checks were not included. The variables analyzed were age, sex, clinical data, and ultrasound and mammography aspects. The devices used are: a Voluson 730 PRO ultrasound machine and a G 600T type mammography machine. Results: At the end of our study, we collected 254 breast pathologies on a number of 382 women, i.e. a frequency of 66.49%. The average age of our patients was 41 years old. The dominant clinical data were mastodynia (41.88%) and mammary nodule (21.70%). On imaging (mammo-ultrasound) the lesions predominated on the left in 36% of cases, bilateral in 28% of cases and in the upper-outer quadrants in 31.5% of cases. Tumor pathologies represented 66.54% of which 45.27% were benign mainly composed of fibro-adenoma (20.88%) and cyst (18.50%), 11.8% of suspected cases and 9.45% of cancers. Non-tumor pathologies represented 33.46%, mainly mastitis (16.14%), galactophoric dilations (11.02%) and abscesses (5.51%). These pathologies were classified in 50.3% in ACR2, 17.75% in ACR3 and 4, and in 14.20% in ACR5. Lymphadenopathy was present in 73.21% of cases.展开更多
为解决传统印刷品图像配准算法精度和速度不能兼得的问题,本文提出了一种基于外轮廓与霍夫变换的印刷品图像配准算法。该算法先对图像进行灰度化、去噪、对比度增强、图像分割完成图像预处理;根据图像分割结果获得印刷区域的外轮廓,计...为解决传统印刷品图像配准算法精度和速度不能兼得的问题,本文提出了一种基于外轮廓与霍夫变换的印刷品图像配准算法。该算法先对图像进行灰度化、去噪、对比度增强、图像分割完成图像预处理;根据图像分割结果获得印刷区域的外轮廓,计算外轮廓形心坐标,得到偏移量,再根据轮廓计算出旋转量;根据偏移量和旋转量对待配准图像进行刚性变换和双线性插值完成图像粗配准;在感兴趣区域(Region of interest,ROI)中限制霍夫变换的动态参数,计算出ROI中的直线,根据直线交点得到3个坐标,根据3个坐标对待配准图像进行仿射变换和双线性插值,完成图像精配准。实验验证表明,本算法配准结果的平均NNC系数达到0.903264,平均配准时间为256.91。展开更多
A computer vision approach through Open AI’s CLIP, a model capable of predicting text-image pairs, is used to create an AI agent for Dixit, a game which requires creative linking between images and text. This paper c...A computer vision approach through Open AI’s CLIP, a model capable of predicting text-image pairs, is used to create an AI agent for Dixit, a game which requires creative linking between images and text. This paper calculates baseline accuracies for both the ability to match the correct image to a hint and the ability to match up with human preferences. A dataset created by previous work on Dixit is used for testing. CLIP is utilized through the comparison of a hint to multiple images, and previous hints, achieving a final accuracy of 0.5011 which surpasses previous results.展开更多
基金Under the auspices of the National Natural Science Foundation of China (No. 40371030), Research Project of Ministry of Construction (No. 20031018)
文摘Researchers in several disciplines and fields agree that the image establishing has both perceptual and affec-tive evaluations.The paper comparatively analyzes the image of landscapes in ancient water towns held by 1619 tourists after the investigations of Zhouzhuang and Tongli of Jiangsu Province.Based on the elements of the image of urban landscapes,the perceptual/cognitive image of tourist landscapes in water towns is developed including path,node,landmark,important courtyard,district and edge.Particularly the important courtyard plays the great role on the per-ceptive image of tourists,however,the perceptual/cognitive images of the district and the edge are obviously weak.Meanwhile,the finding showed that the affective images of tourist landscapes in Zhouzhuang differ with those of Tongli by the method of paired comparison.The main reason is that the affective images are influenced by the impressive scenery spots.Furthermore,the commercialization of streets and pollution of rivers are easy to be functioned negatively.
基金the National Natural Science Foundation of China(No.61877002)Beijing Municipal Commission of Education(PXM2019014213000007).
文摘Image captioning models typically operate with a fixed vocabulary,but captioning is an open-vocabulary problem.Existing work addresses the image captioning of out-of-vocabulary words by labeling it as unknown in a dictionary.In addition,recurrent neural network(RNN)and its variants used in the caption task have become a bottleneck for their generation quality and training time cost.To address these 2 essential problems,a simpler but more effective approach is proposed for generating open-vocabulary caption,long short-term memory(LSTM)unit is replaced with transformer as decoder for better caption quality and less training time.The effectiveness of different word segmentation vocabulary and generation improvement of transformer over LSTM is discussed and it is proved that the improved models achieve state-of-the-art performance for the MSCOCO2014 image captioning tasks over a back-off dictionary baseline model.
文摘Breast pathology is varied, bringing together tumor and non-tumor lesions. Objective: To study the contribution of the ultrasound-mammography pair in the diagnosis of breast pathologies. Materials and Method: This was a retrospective descriptive study, carried out over a period of 3 years (from January 2018 to December 2020) at the Diagnostic Imaging Center (C.I.D) “TERIYA” in BAMAKO. It concerned all patients who came for a mammogram/ultrasound examination of the breast. All women admitted for mammogram or breast ultrasound who were diagnosed with a breast injury during the study period were included. Incomplete records and radiological checks were not included. The variables analyzed were age, sex, clinical data, and ultrasound and mammography aspects. The devices used are: a Voluson 730 PRO ultrasound machine and a G 600T type mammography machine. Results: At the end of our study, we collected 254 breast pathologies on a number of 382 women, i.e. a frequency of 66.49%. The average age of our patients was 41 years old. The dominant clinical data were mastodynia (41.88%) and mammary nodule (21.70%). On imaging (mammo-ultrasound) the lesions predominated on the left in 36% of cases, bilateral in 28% of cases and in the upper-outer quadrants in 31.5% of cases. Tumor pathologies represented 66.54% of which 45.27% were benign mainly composed of fibro-adenoma (20.88%) and cyst (18.50%), 11.8% of suspected cases and 9.45% of cancers. Non-tumor pathologies represented 33.46%, mainly mastitis (16.14%), galactophoric dilations (11.02%) and abscesses (5.51%). These pathologies were classified in 50.3% in ACR2, 17.75% in ACR3 and 4, and in 14.20% in ACR5. Lymphadenopathy was present in 73.21% of cases.
文摘为解决传统印刷品图像配准算法精度和速度不能兼得的问题,本文提出了一种基于外轮廓与霍夫变换的印刷品图像配准算法。该算法先对图像进行灰度化、去噪、对比度增强、图像分割完成图像预处理;根据图像分割结果获得印刷区域的外轮廓,计算外轮廓形心坐标,得到偏移量,再根据轮廓计算出旋转量;根据偏移量和旋转量对待配准图像进行刚性变换和双线性插值完成图像粗配准;在感兴趣区域(Region of interest,ROI)中限制霍夫变换的动态参数,计算出ROI中的直线,根据直线交点得到3个坐标,根据3个坐标对待配准图像进行仿射变换和双线性插值,完成图像精配准。实验验证表明,本算法配准结果的平均NNC系数达到0.903264,平均配准时间为256.91。
文摘A computer vision approach through Open AI’s CLIP, a model capable of predicting text-image pairs, is used to create an AI agent for Dixit, a game which requires creative linking between images and text. This paper calculates baseline accuracies for both the ability to match the correct image to a hint and the ability to match up with human preferences. A dataset created by previous work on Dixit is used for testing. CLIP is utilized through the comparison of a hint to multiple images, and previous hints, achieving a final accuracy of 0.5011 which surpasses previous results.
