Various Applications of Deep Learning .pptx

Applications of Deep Learning.
Various application of deep learning are given in this ppt.

1. Applications of Deep Learning Algorithms/ Architecures Dr. S.Umamaheswari Associate Professor Dept. of Information Technology Anna University MIT Campus. [email protected]

2. How traditional machine learning and deep learning differ?

3. Deep learning architectures Recurrent   neural networks …. LSTM networks. ... Convolutional  neural networks . ... Deep  belief networks. ... Deep  stacking networks.

4. Where can deep learning algorithm/architecture be used?

5. Example: Image Captioning

6. Image impainting

7. Applications of deep learning in Agriculture Why Agriculture? Agriculture is the backbone of any country To support the increasing human population by 2050, the agricultural production must be doubled Its productivity contributes in full filling the basic need of the human society. The productivity is therefore must be smoothen to provide quality and quantity.

8. Yield Prediction e, is of high importance for yield mapping, yield estimation, matching of crop supply with demand, and crop management to increase productivity Water management in agricultural production requires significant efforts and plays a significant role in hydrological, climatological, and agronomical balance Soil management, such as the estimation of soil drying, condition, temperature, and moisture content helps in selecting the crops and improving the quality of produce.

9. Currently our focus Drought, floods, climate change, soil fertility, irrigation and drainage methods, pest, weed etc. affect the agricultural production Weed is unwanted plant that grows in the field. Wide spray of agrochemical degrades the quality of the production and human health Selective spraying or mechanical removal of weed require effective and efficient identification of weed from crop

10. Therefore, for classification of weed from crop, an encoder-decoder network having various layers(Convolution layer with ReLU , Normalization layer and Pooling layer) is employed This paves ways for both selective spraying and mechanical weed removal with decrease in effect of pesticides and labour cost

11. 15 Architecture Preprocessing Encoder Decoder Softmax   Segmented Images RGB Input Images Encoder-Decoder Architecture

12. Core Network Core Network

13. Dataset Description The CWFID dataset is Crop Weed Field Image Dataset It consists of 60 images Also has RGB images, Vegetation segmented masks and crop/weed plant annotations of the same The CWFID is of 88MB Carrot Field Images It consists of 69 Carrot images, 55 Weed images and 54 both Carrot and Weed images The Carrot Field Images are of 982 MB

14. Performance measures   F1 = 2 *   where , Recall = Precision =   true positive : correct positive prediction true negative : correct negative prediction false positive : incorrect positive prediction false negative : incorrect negative prediction  

15.   SegNet 512 SegNet 256 Val_F1 (%) 92 92 Test_F1 (%) 96 93

16.   SegNet 512 SegNet 256 Training Time (sec) 25092.84   11389.65   Evaluate Time (sec) 28.27   12.06 Predict Time (sec) 28.89   12.79   Total Parameters (in thousands) 16,357 4,068   Overall Comparison