diff --git a/data/tuto3-1/vtt/tuto3-activite1-vid1-en.vtt b/data/tuto3-1/vtt/tuto3-activite1-vid1-en.vtt index 80aef9d7e1bf71cd2bfa01fd7d45a2424a06b553..1cd1ea4a81276fae3c5dd64be562757db66d1c98 100644 --- a/data/tuto3-1/vtt/tuto3-activite1-vid1-en.vtt +++ b/data/tuto3-1/vtt/tuto3-activite1-vid1-en.vtt @@ -1,31 +1,55 @@ WEBVTT -Kind: captions -Language: en-GB -00:00:00.000 --> 00:00:02.017 -To create images, A.I. relies on very special Neural Networks.... +00:00:04.528 --> 00:00:07.349 +To create images, artificial intelligence -00:00:02.017 --> 00:00:08.017 -Generative Adversarial Networks, better known as GAN. +00:00:07.349 --> 00:00:09.771 +relies on very special Neural Networks… -00:00:09.017 --> 00:00:12.017 -The GAN is a kind of abracadabra brawler for creating images. +00:00:09.771 --> 00:00:12.261 +Generative Adversarial Networks, -00:00:13.017 --> 00:00:15.017 -GAN consists of two neural networks competing against each other +00:00:12.261 --> 00:00:14.031 +better known as GAN. -00:00:16.017 --> 00:00:19.017 -- the generator , - the discriminator. +00:00:14.031 --> 00:00:16.670 +The GAN is a kind of abracadabra brawler -00:00:20.017 --> 00:00:26.017 -The two neural networks compete against each other and one tries to fool the other. +00:00:16.670 --> 00:00:18.055 +for creating images. -00:00:27.017 --> 00:00:31.017 -The task for the generator is to generate thousands and thousands of new images. +00:00:18.055 --> 00:00:20.729 +GAN consists of two neural networks -00:00:32.017 --> 00:00:36.017 -For the discriminator, it is to know wether the image is generated or not. +00:00:20.729 --> 00:00:22.135 +competing against each other -00:00:37.017 --> 00:00:41.017 -And you, are you a good neural network ? Will you be able to flush out the AI ? +00:00:22.135 --> 00:00:24.621 +the generator, the discriminator. +00:00:24.621 --> 00:00:26.604 +The two neural networks + +00:00:26.604 --> 00:00:28.074 +compete against each other + +00:00:28.074 --> 00:00:30.111 +and one tries to fool the other. + +00:00:30.111 --> 00:00:33.000 +The task for the generator is to generate + +00:00:33.000 --> 00:00:35.525 +thousands and thousands of new images. + +00:00:35.525 --> 00:00:37.830 +For the discriminator, it is to know + +00:00:37.830 --> 00:00:40.194 +whether the image is generated or not. + +00:00:40.194 --> 00:00:42.920 +And you, are you a good neural network? + +00:00:42.920 --> 00:00:45.703 +Will you be able to flush out the AI? diff --git a/data/tuto3-1/vtt/tuto3-activite1-vid2-en.vtt b/data/tuto3-1/vtt/tuto3-activite1-vid2-en.vtt index 7d29e69c7fbbed88e7e4b3796d377e589b64df77..65516b8fe23befb73b258c9eb99b43221ca27b2f 100644 --- a/data/tuto3-1/vtt/tuto3-activite1-vid2-en.vtt +++ b/data/tuto3-1/vtt/tuto3-activite1-vid2-en.vtt @@ -1,43 +1,64 @@ WEBVTT -Kind: captions -Language: en-GB -00:00:00.001 --> 00:00:03.600 -So, are you a good discriminating neural network? +00:00:04.397 --> 00:00:06.150 +So, are you a good -00:00:03.600 --> 00:00:08.750 -Each time the generated image is considered as real by the discriminator network, +00:00:06.150 --> 00:00:08.096 +discriminating neural network? -00:00:08.750 --> 00:00:13.634 -the generative network reinforces its parameters and progressively improves. +00:00:08.096 --> 00:00:10.670 +Each time the generated image is considered -00:00:13.634 --> 00:00:17.817 -Some elements of the image can particularly betray a generative AI: +00:00:10.670 --> 00:00:12.559 +as real by the discriminator network, -00:00:17.817 --> 00:00:19.800 +00:00:12.559 --> 00:00:14.619 +the generative network reinforces + +00:00:14.619 --> 00:00:17.293 +its parameters and progressively improves. + +00:00:17.293 --> 00:00:19.230 +Some elements of the image + +00:00:19.230 --> 00:00:21.920 +can particularly betray a generative AI: + +00:00:21.920 --> 00:00:23.807 the background of the image, -00:00:19.800 --> 00:00:21.767 +00:00:23.807 --> 00:00:24.865 the teeth, -00:00:21.767 --> 00:00:23.817 +00:00:24.865 --> 00:00:27.645 the asymmetry of the face or eyes, -00:00:23.817 --> 00:00:25.884 +00:00:27.645 --> 00:00:30.000 unexpected blurred areas, -00:00:25.884 --> 00:00:30.700 +00:00:30.000 --> 00:00:31.840 a hair a bit strange. -00:00:30.700 --> 00:00:34.900 -If you look carefully, the images generated by GANs are still recognisable, +00:00:31.840 --> 00:00:33.524 +If you look carefully, + +00:00:33.524 --> 00:00:35.565 +the images generated by GANs -00:00:34.900 --> 00:00:40.717 -but these bugs may well be corrected in the years to come and it will be increasingly difficult to distinguish them! +00:00:35.565 --> 00:00:37.392 +are still recognizable, -00:00:40.717 --> 00:00:43.017 +00:00:37.392 --> 00:00:40.000 +but these bugs may well be corrected + +00:00:40.000 --> 00:00:41.671 +in the years to come and it will be + +00:00:41.671 --> 00:00:44.000 +increasingly difficult to distinguish them! + +00:00:44.000 --> 00:00:45.627 Want to try again? -00:00:43.134 --> 00:00:45.617 +00:00:45.627 --> 00:00:48.142 It's up to you! - diff --git a/data/tuto3-1/vtt/tuto3-activite1-vid3-en.vtt b/data/tuto3-1/vtt/tuto3-activite1-vid3-en.vtt index 1b98b2bb0e2053fffe3cc613355219e714cd0d48..2dd26cf1f398fedd1e4f232b69481c8be68d211a 100644 --- a/data/tuto3-1/vtt/tuto3-activite1-vid3-en.vtt +++ b/data/tuto3-1/vtt/tuto3-activite1-vid3-en.vtt @@ -1,35 +1,55 @@ WEBVTT -Kind: captions -Language: en-GB -00:00:00.000 --> 00:00:02.120 -This new type of neural network raises many questions +00:00:04.159 --> 00:00:06.160 +This type of new neural network -00:00:02.120 --> 00:00:08.600 -... as these GANs exist for all media: -text, music, animated images or videos +00:00:06.160 --> 00:00:07.603 +raises many questions -00:00:08.600 --> 00:00:10.680 +00:00:07.603 --> 00:00:09.850 +as these GANs exist for all media: + +00:00:09.850 --> 00:00:13.509 +text, music, animated images or videos. + +00:00:13.509 --> 00:00:16.084 So, real or fake? -00:00:10.680 --> 00:00:16.700 -It is becoming essential to learn to recognize whether -an image is real or machine generated. +00:00:16.084 --> 00:00:17.930 +It is becoming essential to learn + +00:00:17.930 --> 00:00:19.990 +to recognize whether an image is real + +00:00:19.990 --> 00:00:21.281 +or machine generated. + +00:00:21.281 --> 00:00:22.969 +But how long before the details -00:00:16.700 --> 00:00:22.680 -But how long before the details are no longer visible? +00:00:22.969 --> 00:00:24.485 +are no longer visible? -00:00:22.680 --> 00:00:28.800 -Are all types of content affected by these generative possibilities? +00:00:24.485 --> 00:00:26.670 +Are all types of content affected -00:00:28.800 --> 00:00:32.160 +00:00:26.670 --> 00:00:28.629 +by these generative possibilities? + +00:00:28.629 --> 00:00:30.648 Who is the author of these images? -00:00:32.160 --> 00:00:38.880 +00:00:30.648 --> 00:00:32.770 The person who created the programme? -The programme that created images that did not exist? -00:00:38.880 --> 00:00:44.840 +00:00:32.770 --> 00:00:34.460 +The programme that created images + +00:00:34.460 --> 00:00:35.703 +that did not exist? + +00:00:35.703 --> 00:00:37.582 And if we don't know if it's real, -is there an annihilation of fiction? +00:00:37.582 --> 00:00:40.309 +is there an annihilation of fiction?