Auto Topic: cnn
auto_cnn | topic
Coverage Score
1
Mentioned Chunks
15
Mentioned Docs
1
Required Dimensions
definitionpros_cons
Covered Dimensions
definitionpros_cons
Keywords
cnn
Relations
| Source | Type | Target | W |
|---|---|---|---|
| Auto Topic: cnn | CO_OCCURS | Auto Topic: convolutional | 7 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: dimension | 5 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: pixels | 5 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: convolution | 4 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: kernel | 4 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: stride | 4 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: window | 4 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: spatial | 3 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: cnns | 3 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: detector | 3 |
| Auto Topic: cnn | CO_OCCURS | Auto Topic: windows | 3 |
Evidence Chunks
| Source | Confidence | Mentions | Snippet |
|---|---|---|---|
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.57 | 2 | ... nd spatial invariance—we can develop models that have far fewer parameters and can learn much more quickly. A convolutional neural network (CNN)is one that contains spatially local connections, Convolutional neural network (CNN) at least in the early layers, and has |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.57 | 2 | rameters and can learn much more quickly. A convolutional neural network (CNN)is one that contains spatially local connections, Convolutional neural network (CNN) at least in the early layers, and has patterns of weights that are replicated across the units in each layer. A patte ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.57 | 2 | ... ing at a small sliding window onto the largerSliding window image—a rectangle. At each spot, we classify what we see in the window, using a CNN classifier. We then take the high-scoring classifications—a cat over here and a dog over there—and ignore the other windows. After some wo ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... dn’t necessarily construct the weight matrix explicitly—it is Section 22.3 Convolutional Networks 813 Figure 22.5 The first two layers of a CNN for a 1D image with a kernel size l =3 and a stride s =1. Padding is added at the left and right ends in order to keep the hidden layers ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... se models, the receptive field of a neuron is the portion of the sensory input that can Receptive field affect that neuron’s activation. In a CNN, the receptive field of a unit in the first hidden layer is small—just the size of the kernel, i.e., l pixels. In the deeper layers of the ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... ssify the image into one of c categories, then the final layer of the network will be a softmax with c output units. The early layers of the CNN are image-sized, so somewhere in between there must be significant reductions in layer size. Convolution layers and pooling layers with s ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... ds has throughput equivalent to about ten million laptops. Taking advantage of these capabilities is essential if one is train- ing a large CNN on a large database of images. Thus, it is common to process not one image at a time but many images in parallel; as we will see in Sect ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... an performance) and 87% in top-1 accuracy by 2019. The primary reason for this success seems to be that the features that are being used by CNN classifiers are learned from data, not hand-crafted by a researcher; this ensures that the features are actually useful for classification ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... multiple levels, and is doing that by learning from the data rather than having the patterns given to it by a programmer. While training a CNN “out of the box” does sometimes work, it helps to know a few practical techniques. One of the most important is data set augmentation, i ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... dle this well, classifying an image as “cat” accurately even if few pixels actually lie on the cat. There are two reasons for this. First, CNN-based classifiers are good at ignoring patterns that aren’t discriminative. Second, patterns that lie off the object might be discriminati ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... th.) We still need to choose the width and height of the rectangles. • Build a classifier for windows: We already know how to do this with a CNN. • Decide which windows to look at:Out of all possible windows, we want to select ones that are likely to have interesting objects in th ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... type of box. Now any box with a good enough objectness score is called a region of interest (ROI), and must be checked by a classifier. But CNN classifiers prefer images of fixed size, and the boxes that pass the objectness test will differ in size and shape. We can’t make the boxe ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... n the abstract style (see Figure 27.26). The key insight to solving this problem is that if we examine a deep convolutional neural network (CNN) that has been trained to do object recognition (say, on ImageNet), we find that the early layers tend to represent the style of a pictur ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... 789 clutter (in data association), 657 CNF (conjunctive normal form), 244, 244–245, 265, 317–318 CNLP (conditional nonlinear planning), 401 CNN (convolutional neural network), 811, 1003 co-NP, 1076 co-NP-complete, 240, 1076 coalition, 616 coalition structure, 616 coalition struct ... |
textbook Artificial-Intelligence-A-Modern-Approach-4th-Edition.pdf | 0.55 | 1 | ... ation, 140, 159 CONVINCE (Bayesian expert system), 472 convolution, 997 convolution (in neural networks), 811 convolutional neural network (CNN), 811, 1003 Conway, D., 738, 1091 Cook, P. J., 1051, 1095 Cook, S. A., 27, 266, 267, 1080, 1091 Cooper, G., 475, 798, 1091 Cooper, M. C. ... |