How to build an object detection model from scratch?
Answers from top 13 papers
More filters
| Papers (13) | Insight |
|---|---|
07 Oct 2010 | Our approach could lend itself nicely to real-time object tracking and detection. |
| We propose a new method of background modeling for object detection. | |
23 Jun 2014 24 Citations | We show how this framework can be used also for detection rather than tracking which allows for automatic system initialization and for incorporating pose estimates obtained from independent object part detectors. |
07 Jun 2015 | Our approach achieves the state-of-the-art performance for both object proposal and object detection tasks, comparing to previous object proposal methods. |
29 Dec 2011 | In this paper we propose using a pictorial structure model for object detection, and modify it to better perform in a drawing setting as opposed to photographs. |
23 Jun 2013 | In this paper, we present the Efficient Maximum Appearance Search (EMAS) model which is an order of magnitude faster than the existing state-of-the-art large-scale object detection approaches, while maintaining comparable accuracy. |
| In particular, we show how to create new detectors on the fly using large-scale internet image databases, thus allowing a user to choose among thousands of available categories to build a detection system suitable for the particular robotic application. | |
128 Citations | Our model advances the state of the art for object detection by improving the accuracy of template matching. |
07 Oct 2012 | We propose a new model for object detection that is based on set representations of the contextual elements. |
| Experimental results show that the model can achieve high accuracy in predicting the performance of object detection. | |
20 Jun 2011 | Our experimental results on the challenging object localization and detection task demonstrate that the proposed model substantially improves the accuracy of the standard single object detectors that ignore hierarchical structure altogether. |
16 Jun 2019 | Furthermore, our model is the first one that fuses visual semantic with 3D object detection. |
01 Sep 2009 181 Citations | We demonstrate the robustness of our model on object detection, viewpoint classification and synthesis tasks. |
Related Questions
What are the state of the art models for object detection?5 answersThe state-of-the-art models for object detection include Convolutional Neural Networks (CNNs), deep learning-based object detection algorithms, and Deep Learning (DL) models. Specifically, YOLOv5 has been used for real-time mask detection, and various deep learning-based object detection algorithms have been investigated for localizing prostate cancer metastatic lesions in PET images. YOLOv5 has also been used in the context of smart farming for object detection with limited labeled training data. Additionally, a list of datasets and their best models has been provided for advancing the state-of-the-art in object detection.
What ai models are best for object detection?5 answersIn the field of object detection, several AI models have been studied and compared for their accuracy and speed. Based on the abstracts provided, the following models have been found to be effective for object detection:
1. YOLOv7: This model achieves high accuracy and fast inference speed, making it suitable for real-time object detection. It outperforms other models like Cascade-Mask R-CNN in terms of accuracy and speed.2. Inception v3: This pre-trained model has shown superior performance in terms of accuracy and speed compared to other models like ResNet-50, DenseNet-121, and SqueezeNet. It can handle images of different sizes, making it suitable for real-world communication when integrated with IoT-based devices.3. Fine-tuned CNN-based models: Fine-tuning a CNN-based model, such as the one trained on the detection of paper boxes, has been found to be effective in minimizing the number of images required for model fitting while ensuring reliable and robust performances.4. YOLOv5: This model, when combined with Active Learning (AL) techniques, has shown promising results in enhancing the efficiency of object detection models trained with a small dataset. AL helps reduce the amount of data and time required for training, making it suitable when labeled training data is limited.These models offer different advantages and can be chosen based on specific requirements and constraints.
Can we train the model from scratch without any changes in architecture?5 answersSparse training from scratch in deep reinforcement learning (DRL) has not been well explored. Existing methods mainly adopt the knowledge distillation-based approach by iteratively training a dense network. However, a novel sparse DRL training framework called "the Rigged Reinforcement Learning Lottery" (RLx2) has been proposed. RLx2 is capable of training a sparse DRL model based entirely on a sparse network. It introduces a novel multi-step TD target mechanism with a dynamic-capacity replay buffer to achieve robust value learning and efficient topology exploration in sparse models. RLx2 achieves state-of-the-art sparse training performance in several tasks, showing significant model compression with minimal performance degradation and reduced FLOPs for training and inference.
What is better when training object detection model on anew dataset, start from scrach or transfer learning?5 answersTransfer learning is generally better when training an object detection model on a new dataset compared to starting from scratch. Transfer learning allows the model to leverage pre-trained weights from similar source datasets, improving feature extraction and convergence speed. It also helps to mitigate the domain gap between standard object detection datasets and artistic imagery, which can improve performance on artistic data. Additionally, using a dataset-agnostic template for object detection trainings, consisting of pre-trained models and a robust training pipeline, provides a strong baseline on a wide range of datasets. Transfer learning-based approaches, such as YOLOv3, have shown superior object detection accuracy compared to traditional methods. Furthermore, contrastive learning with style transfer has been proposed as a solution for cross-domain object detection, achieving state-of-the-art performance.
How to prepare a dataset for object detection?8 answers
Is object detection deep learning?4 answers