1 00:00:00,00 --> 00:00:06,02 (bright upbeat music) 2 00:00:06,02 --> 00:00:08,08 - [Sudha] I have a fun challenge for you. 3 00:00:08,08 --> 00:00:12,09 Take a photo of your environment as seen from your camera. 4 00:00:12,09 --> 00:00:15,08 It can be a picture of a production line 5 00:00:15,08 --> 00:00:17,06 or a factory 6 00:00:17,06 --> 00:00:19,03 or a warehouse environment 7 00:00:19,03 --> 00:00:23,03 or a roadside view from a security camera. 8 00:00:23,03 --> 00:00:26,04 The camera is the edge AI device 9 00:00:26,04 --> 00:00:30,01 and the photo is going to be the data. 10 00:00:30,01 --> 00:00:32,05 Follow the steps from the demo. 11 00:00:32,05 --> 00:00:35,07 Feel free to refer to the handout for the steps. 12 00:00:35,07 --> 00:00:37,01 Input your image 13 00:00:37,01 --> 00:00:40,09 into the Azure AI object detection interface 14 00:00:40,09 --> 00:00:43,07 and see what objects are detected. 15 00:00:43,07 --> 00:00:46,03 Remember, we click Browse for a file, 16 00:00:46,03 --> 00:00:47,09 found the file, 17 00:00:47,09 --> 00:00:50,09 and then we input the file. 18 00:00:50,09 --> 00:00:53,00 You can do that with your own images, 19 00:00:53,00 --> 00:00:56,09 and you can try this with more than one image also. 20 00:00:56,09 --> 00:01:00,09 Write down what objects are detected 21 00:01:00,09 --> 00:01:04,01 and what objects you expect to be detected 22 00:01:04,01 --> 00:01:05,06 but are not detected. 23 00:01:05,06 --> 00:01:09,00 For example, you've seen this in the demo. 24 00:01:09,00 --> 00:01:15,05 This image shows two people detected. 25 00:01:15,05 --> 00:01:20,00 Actually, it shows one more person detected in the back. 26 00:01:20,00 --> 00:01:21,09 It does not detect this person. 27 00:01:21,09 --> 00:01:26,00 So I would expect all four humans to be detected, 28 00:01:26,00 --> 00:01:28,05 but this person is missing. 29 00:01:28,05 --> 00:01:30,06 If you want, you can adjust the threshold 30 00:01:30,06 --> 00:01:32,08 to see does that change anything? 31 00:01:32,08 --> 00:01:35,03 It doesn't seem to change much. 32 00:01:35,03 --> 00:01:37,07 And if you want the objects to be detected, 33 00:01:37,07 --> 00:01:40,08 that is not being detected in this picture. 34 00:01:40,08 --> 00:01:44,01 So your exercise challenge 35 00:01:44,01 --> 00:01:47,00 is to write down what objects are detected 36 00:01:47,00 --> 00:01:49,08 and what objects are not detected. 37 00:01:49,08 --> 00:01:51,03 After adjusting the threshold, 38 00:01:51,03 --> 00:01:55,03 write down what threshold are you selecting that makes sense 39 00:01:55,03 --> 00:01:57,06 to show the degree of confidence 40 00:01:57,06 --> 00:02:01,09 with which you want this AI to work for your use case. 41 00:02:01,09 --> 00:02:04,08 Does this meet your customer need for your work environment? 42 00:02:04,08 --> 00:02:09,06 For safety or identifying defects or any other use case, 43 00:02:09,06 --> 00:02:13,08 you would like to use computer vision for object detection. 44 00:02:13,08 --> 00:02:19,02 Finally, what would you change to meet your needs? 45 00:02:19,02 --> 00:02:22,01 Would you change the threshold? 46 00:02:22,01 --> 00:02:26,01 Would you collect a new set of images of your environment 47 00:02:26,01 --> 00:02:29,01 and train this pre-trained model? 48 00:02:29,01 --> 00:02:31,06 And I already showed you here in the demo 49 00:02:31,06 --> 00:02:34,00 that if you to want use your own images, 50 00:02:34,00 --> 00:02:36,02 you have to create an Azure account, 51 00:02:36,02 --> 00:02:39,01 and then you can go about using the API. 52 00:02:39,01 --> 00:02:41,02 And you have all the details here, 53 00:02:41,02 --> 00:02:43,02 but it is going to cost you token cost. 54 00:02:43,02 --> 00:02:45,08 I want you to be aware of that. 55 00:02:45,08 --> 00:02:47,06 Feel free to ask any questions 56 00:02:47,06 --> 00:02:49,09 in the question tab of the course, 57 00:02:49,09 --> 00:02:52,08 or come to my livestream, "SudhaLive," 58 00:02:52,08 --> 00:02:54,04 that I host monthly, 59 00:02:54,04 --> 00:02:56,03 and I talk to all my students. 60 00:02:56,03 --> 00:02:59,07 So you can scroll down at the bottom of any of my courses 61 00:02:59,07 --> 00:03:01,06 and you will see the "SudhaLive" link. 62 00:03:01,06 --> 00:03:04,09 And I host that once a month right here on LinkedIn. 63 00:03:04,09 --> 00:03:08,01 And you can also tag my learner community 64 00:03:08,01 --> 00:03:11,07 at Business School of AI if you have a question. 65 00:03:11,07 --> 00:03:15,00 Next, I'm going to show you the solution. 66 00:03:15,00 --> 00:03:17,00 Remember, I'm here to help.