AI-based object detection latest trends in remote sensing, multimedia and agriculture applications

AI has the ability to detect and classify tumors by analyzing brain imaging techniques, such as Magnetic Resonance Imaging (MRI). AI algorithms can help determine the size, location, class, and aggressiveness of tumors. This helps physicians make a more accurate diagnosis and treatment plan, and helps patients better understand their health. P.Ab and S.J.M.J. provided advice on machine learning analysis and provided computational resources. Reviewed all specimens for histological and molecular pathology and contributed to manuscript writing.

• American universities have a long history of welcoming foreign engineering graduate students who go on to have highly successful decades-long careers in the U.S.
• Recently, AI-based image analysis models outperformed human labor in terms of the time consumed and accuracy7.
• In application, by presenting a heatmap, it provides context and evidence demonstrating how the diagnosis was achieved.
• The first step is the design of the test programme and the presentation of the model parameters.
• The algorithm designed a deep belief network structure and conducted experiments on feature extraction, and finally achieved an accuracy of 77%.

DenseNet is a deep learning neural network architecture where each layer is connected to all previous layers and information can flow fully and efficiently through the network. This feature allows the DenseNet network to better learn and understand the details and structure of an image. Due to the dense connectivity, the DenseNet network enables feature reuse, which improves the algorithm’s feature representation and learning efficiency. In addition, the DenseNet network structure is simple and has a small number of parameters, which solves the problem of gradient vanishing that commonly exists in neural network models. Therefore, the study selected DenseNet as the object of research, and the DenseNet structure is shown in Fig. Most object detection deep neural network models were proficient with objects of varying sizes.

Now, though, new research claims that locally run bots using specially trained image-recognition models can match human-level performance in this style of CAPTCHA, achieving a 100 percent success rate despite being decidedly not human. As part of the study, CNN and CNN-based transfer learning models such as InceptionV3, EfficientNetB4, VGG19 were trained on open-source shared brain tumor patients. The comparison of the brain tumor studies with the literature is shown in Table 4. The metrics tracked during training are accuracy, as well as precision and recall.

Could AI-powered image recognition be a game changer for Japan’s scallop farming industry?

The experiment findings denoted that the model had better recognition performance than existing common facial recognition algorithms. The accuracy, precision, recall, and F1 value indicators reached 99.2%, 96%, 98%, and 96%, respectively11. The method could achieve multi-level classification and deep extraction of image features.

This second set is trained to detect 12 types of pathological findings (e.g., pneumonia, fracture, pneumothorax, etc.) and is then evaluated in a binary task of predicting whether or not there are any pathological findings present (Fig. 1b). In this task, Seyyed-Kalantari et al. identified an underdiagnosis bias for underserved populations, where, for instance, Black patients were more likely to have a false negative result by the AI algorithm compared to white patients1. Figure 6 compares the heatmaps generated by the proposed AIDA, with those produced by the Base and CNorm for the source (a) and target (b and c) domains of the Pleural dataset. While all samples were diagnosed as “Malignant”, both the Base and CNorm approaches incorrectly classified the slides as “Benign”. However, by applying AIDA, the majority of patches were accurately classified as “Malignant”, ultimately leading to the correct classification of the entire slide as “Malignant”. Similar to the observations made in the context of ovarian cancer, a notable alignment is observed between the tumor annotations provided by the pathologist and the heatmap generated by the AIDA method for pleural cancer.

2 Automated Chilli disease detection

4, the transfer learning models VGG, InceptionV3, and EfficientNetB4 and the models built with CNN have distinctive features. The confusion matrix of the study on the classification of glioma, meningioma, non-tumor normal patients, pituitary tumor patients in the dataset by tumor type is shown in Fig. This architecture stands as a notable CNN model introduced by24, which builds upon its predecessor, the AlexNet model.

Like a human, AGI could potentially understand any intellectual task, think abstractly, learn from its experiences, and use that knowledge to solve new problems. Essentially, we’re talking about a system or machine capable of common sense, which is currently unachievable with any available AI. Reinforcement learning is also used in research, where it can help teach autonomous robots the optimal way to behave in real-world environments. Robots learning to navigate new environments they haven’t ingested data on — like maneuvering around surprise obstacles — is an example of more advanced ML that can be considered AI. AI has a slew of possible applications, many of which are now widely available in everyday life. At the consumer level, this potential includes the newly revamped Google Search, wearables, and even vacuums.

For the DICOM-based evaluation, we use the same list of images as the original MXR test set but extract the pixel data from the corresponding DICOM files instead of using the preprocessed JPEG files. We restrict this evaluation to MXR because the original DICOM files are not publicly available for CXP. When evaluating the AI models on the DICOM images, we first extract and process the pixel data according to the DICOM Standard58 using code based on the pydicom library59. This processing includes using the default windowing parameters in the DICOM header, as would be done by standard DICOM viewers.

It also offers advanced search tools with certain criteria, or you can create your own criteria to meet specific requirements. Mylio Photos is an advanced media management application that uses artificial intelligence to greatly enhance the organization, searching, and editing of digital memories. The application offers AI smart tags and a sophisticated search functionality, enabling users to locate images efficiently across different devices without manual searches. The AI recognizes content and context, simplifying the discovery process through an intuitive interface that includes calendar views, custom tags, and geographic data.

Dataset descriptions

As ECGs have transitioned from analog to digital, automated computer analysis has gained traction and success in diagnoses of medical conditions (Willems et al., 1987; Schlapfer and Wellens, 2017). Deep learning methods have shown excellent diagnostic performance on classifying ECG diagnoses using signal data, even surpassing individual cardiologist performance in some studies. Other studies used signal data from 12-lead ECGs with excellent results in arrhythmia classification (Baek et al., 2021).

The improved RetinaNet utilizes rotating frames to locate and identify equipment, circumventing the limitations of conventional framing and reducing overlap, thereby achieving more precise detection outcomes. 3 reveals that the NL-means and wavelet transform denoising effects are somewhat inferior compared to Dn-CNN, with more residual noise remaining after NL-means processing and more severe image distortion. The average PSNR for NL-means, wavelet transform, Dn-CNN, and DeDn-CNN are 33.47, 34.82, 38.25, and 40.33, respectively, which further demonstrates that DeDn-CNN is more effective at removing noise from infrared images. Anyone who has been surfing the web for a while is probably used to clicking through a CAPTCHA grid of street images, identifying everyday objects to prove that they’re a human and not an automated bot.

The system learns to analyze the game and make moves, learning solely from the rewards it receives. It can eventually play by itself and learn to achieve a high score without human intervention. This common technique for teaching AI systems uses annotated data or data labeled and categorized by humans. A major function of AI in consumer products is personalization, whether for targeted ads or biometric security. This is why your phone can distinguish your face from someone else’s when you’re unlocking it with Face ID, for example — it’s learned what yours looks like by referencing billions of other people’s faces and matching specific data points.

This highlights the proposed model’s effectiveness in accurately detecting tunnel face lithology. This implies that the compressive strength of rocks needs to be adjusted according to their weathering degree. The application of correction coefficients is a process of adjusting the original compressive strength of the rock based on its weathering condition to obtain a rock strength value that better reflects the actual conditions.

Object detection and recognition applications in agriculture using AI

● Weakly supervised object detection models aim to detect many non-annotated corresponding objects using a small set of fully annotated images. Therefore, using many annotated and labeled pictures with target objects and bounding boxes to train the network to achieve high effectiveness efficiently is an essential issue for future research. Deep CNN plannings generate hierarchy feature maps due to pooling and subsampling operations, resulting in changed layers of feature maps with differing 3D resolutions. As is generally known, the feature maps of the early-layer feature maps have a higher resolution and signify smaller response fields. They also lack high-level semantic information, which is necessary for object detection.

The improved RetinaNet’s recognition results are fed into the DeeplabV3 + model to further segment structures prone to thermal faults. The accuracy of component recognition in this paper achieved 87.23%, 86.54%, and 90.91%, with respective false alarm rates of 7.50%, 8.20%, and 7.89%. The Inception architecture is an architecture used in the field of deep learning and CNN. It is designed to perform feature extraction and classification tasks more efficiently.

Without controlling for the difficulty of images used for evaluation, it’s hard to objectively assess progress toward human-level performance, to cover the range of human abilities, and to increase the challenge posed by a dataset. To address these unmet needs, we developed OrgaExtractor a DL-based organoid image analysis algorithm. OrgaExtractor was designed to overcome the current inefficiency in analyzing organoid images.

Google, Facebook, Microsoft, Apple and Pinterest are among the many companies investing significant resources and research into image recognition and related applications. Privacy concerns over image recognition and similar technologies are controversial, as these companies can pull a large volume of data from user photos uploaded to their social media platforms. Image recognition is used to perform many machine-based visual tasks, such as labeling the content of images with meta tags, performing image content search and guiding autonomous robots, self-driving cars and accident-avoidance systems. Once the pre-trained CNN has been fine-tuned on the retail company’s product images, it can be used to generate embeddings for each product. These embeddings represent the unique features of each product that the model has learned to recognize.

For the source domain of the Ovarian and Pleural datasets, the patches were extracted from the pathologist-annotated areas (containing tumor tissue) while we randomly extracted patches for the target domain. This random extraction approach was adopted for the target domain due to the consideration of target data as unsupervised, with the assumption that no prior information is available ChatGPT for this dataset. For the Bladder dataset, patches were extracted randomly from both the source and target datasets because pathologists’ annotations were not accessible. The patches were then resized to 512 × 512 pixels, ensuring a standardized magnification of 20X. We employed a three-fold cross-validation strategy to evaluate the performance of the model on the source domain.

With Adobe Scan, the mundane task of scanning becomes a gateway to efficient and organized digital documentation. Users can capture images of leaves, flowers, or even entire plants, and PlantSnap provides detailed information about the identified species. Beyond simple identification, it offers insights into care tips, habitat details, and more, making it a valuable tool for those keen on exploring and understanding the natural world. Seeing AI can identify and describe objects, read text aloud, and even recognize people’s faces. Its versatility makes it an indispensable tool, enhancing accessibility and independence for those with visual challenges. By combining the power of AI with a commitment to inclusivity, Microsoft Seeing AI exemplifies the positive impact of technology on people’s lives.

In-depth analysis of the network structure, advantages, disadvantages, and applicable scenarios of various algorithms, we compare the analysis of standard data sets and experimental results of different related algorithms on mainstream data sets. Finally, this study summarizes some application areas of object detection to comprehensively understand and analyze its future development trend. The rapid development of deep learning has increased the feasibility of improving various classical object detection algorithms in many ways. The improved scheme corresponding to the algorithm detection process is shown in Figure 4.

The Ubisa-Shorapani (F3) project route on the E60 highway in Georgia has a total length of 13.04 km, with a design speed of 100 km/h. The project features a two-way, four-lane cement concrete pavement, with a road width of 27.6 m and a lane width of 3.75 m, totaling 7.5 m for one-way lanes. This case study was applied to three tunnels in this construction area, where the geological conditions are complex. According to the geological survey report during the tunnel design phase, the primary rock types in this area are porphyrite and granite, with the strength of these rocks identified as 125 MPa and 175 MPa, respectively. Therefore, this experiment was conducted based on database data of porphyrite and granite, along with some on-site image data. The models were trained on the same batch database and evaluated using several metrics.

Through image processing, the structure and potential weaknesses of the surrounding rock can be clearly identified, providing important reference information for subsequent construction. In summary, integrating the ResNet-18 model for weathering degree determination complements the lithology segmentation model, forming a comprehensive framework for analyzing tunnel face images. This dual-model approach enhances the overall accuracy and efficiency of rock strength assessment in tunnel construction.

Based on the most correlated parameter, frequent image analysis enables the estimation of actual cell numbers in a non-invasive manner in real time and provides opportunity to culture organoid samples continuously (Fig. 1e). While there are challenges to implementing transfer learning in different business domains, such as finding and adapting the right pre-trained models to the specific domain and dataset, the benefits are significant. By leveraging transfer learning, businesses can improve the accuracy and efficiency of their computer vision models, leading to better customer experiences and increased revenue.

We used high-performance DSLR cameras or high-resolution smartphones for photography, adjusting camera parameters to account for the low light and high dust environments typical in tunnels. The optimal time for capturing images is usually after blasting when the dust has settled and before the commencement of preliminary support work, as shown in Fig. This approach effectively captures high-quality images of the tunnel face, providing an accurate data foundation for subsequent deep learning analysis.

An object detection-based few-shot learning approach for multimedia quality assessment, Multimedia Systems (Springer), 1–14. ● Region-specific detectors tend to perform better, achieving higher detection accuracy on predefined datasets. Therefore, developing a general object detector that can detect multi-domain objects without prior knowledge is a fundamental research direction in the future. ● Video object detection has problems such as uneven moving targets, tiny targets, truncation, and occlusion, and it isn’t easy to achieve high precision and high efficiency.

MaskR-CNN, proposed by He et al. (2017) is a Faster R-CNN extension that uses the ResNet-101-FPN backbone network. Multi-task loss is combined with segmentation branch loss, arrangement, and bounding box regression ai based image recognition loss in Mask R-CNN. A Mask network branch for RoI calculation and division is added to the object classification and bounding box regression to enable real-time object identification and instance segmentation.

CNNs, as a typical deep learning method, are widely used in IR and classification. Google Perception Net, Visual Geometry Group (VGG), Residual Neural Network, and DenseNet are all developed from CNNs22,23,24. CNN is a deeply supervised machine learning model with strong adaptability, especially good at mining local features of data, extracting global features and carrying out classification processing (Fig. 1). At the early stage of the training of the CNN, the feature map will be extracted to the edge and texture information of the image. As the network goes deeper, the information with more semantic information level will be extracted, and after several feature extraction operations, the CNN will complete the recognition task of the whole image.

However, current CNN-based sports image classification faces challenges such as limited sport categories (typically fewer than 10) and issues with long prediction times and insufficient accuracy. You can foun additiona information about ai customer service and artificial intelligence and NLP. To improve the accuracy of stroke imaging diagnosis, Hou et al. designed a deep learning based 3D residual IR model for stroke lesions in medical imaging. The model combined attention mechanisms and residual networks, and utilized 3D convolutional kernels to learn continuous information between image sequences.

A unique squeeze-and-excitation-based convolutional neural network (SECNN) model outperformed the rest, obtaining 98.63% accuracy without augmentation and 99.12% with augmentation, respectively (Table 6). The downstream tumor subtype classifier relies on the tumor areas of the tissues. Given that the manual annotation of all slides by pathologists is tedious and time-consuming, we first trained a deep learning model to identify the tumor areas of the slides automatically (Supplementary Fig. 7). To train the model, we utilized 27 slides that were annotated by a board-certified pathologist. First, we split the slides into training (51.8%), validation (22.2%), and testing (26%) sets.

The Base and AIDA differ in terms of model complexity, with AIDA being slightly more intricate than Base. This difference in complexity contributes to variations in their training times. However, during inference, where both networks leverage the same backbone, feature extraction, and multiple instance learning processes require a comparable amount of time for both models.

In addition, many researchers have used deep learning to detect and recognize remote sensing image targets, and have achieved good results and achieved many breakthroughs (Krizhevsky et al., 2017). Mnih and Hinton (2010) used two datasets of remote sensing images to conduct research on deep learning technology. They extracted road features from images for training and achieved good experimental results. The algorithm designed a deep belief network structure and conducted experiments on feature extraction, and finally achieved an accuracy of 77%.

Design of IR and classification algorithm based on improved densenet

Schwarz is currently training the AI model for visual inspection on a new line. Schwarz manually retrains the model until it recognizes faults just as reliably as it does flawless parts. “Stators let us exploit the potential of generative AI particularly well,” Beggel says.

Furthermore, OrgaExtractor performs segmentation tasks precisely within three seconds per image and provides detailed information regarding each organoid in an image. The ability to analyze a single organoid, even in images containing multiple organoids, allows researchers to avoid capturing each organoid, leading to a more efficient examination process. OrgaExtractor trained with a colon organoid image set can recognize same colon organoid sample from different sets of images with an average DSC of over 0.83 (Supplementary Fig. S3). As OrgaExtractor recognizes organoids in a single image, its organoid measurements are comparable to manual measurements in multiple images (Fig. 2).

• To this end, this work introduces big data mining technology to explore educators’ teaching characteristics and behaviors that affect the quality of online courses.
• This step could include retraining the models with fresh data, modifying the features or parameters, or even developing new models to meet new demands.
• Contemporary TSM methods often combine semantic information with various weighting, regularization strategies, and NLP techniques29.

AI-powered photo organizers use machine learning algorithms to automatically tag, sort, and categorize photos based on their content, date, location, and other factors. These intelligent tools are becoming essential in the digital age, allowing us to quickly locate specific photos and share them with ease. By following a systematic dataset curation flow, bulleted below and depicted in Fig.

Further details regarding the proposed network are provided in the Methods section. A The patches are first extracted from the WSIs in both the source ChatGPT App and target domains. B Through FFT-Enhancer, patch colors from the source domain are adjusted to look more like patches from the target domain.