Jump to content
  • Neural network trained to turn any video into slow motion


    DeZire

    The programmer under the nickname MortenHannemose posted on Github the algorithm, thanks to which you can get a quality slow-SloMo-video movies with any framerate. A neural network is used to draw the missing images. Anyone can try the program.

    Neural network trained to turn any video into slow motion_1.gif

    For the plugin to work, you need to install the Python environment and connect the bundled trained library. The essence of the method is to generate the missing frames in order to slow down the video. The neural network takes two adjacent frames and draws an intermediate one.

    The result is a smooth video. The capabilities of the algorithm are superior to the program After Effect, Sepconv, and analogs. The processing speed will depend on the processor power and resolution of the original frame.

    By default, the resolution is limited to 320 pixels, but this can be circumvented by changing just one line of code. Now the Creator of the script continues its active testing and is ready to listen to any suggestions and suggestions from users.



    User Feedback

    Recommended Comments

    There are no comments to display.



    Guest
    Add a comment...

    ×   Pasted as rich text.   Paste as plain text instead

      Only 75 emoji are allowed.

    ×   Your link has been automatically embedded.   Display as a link instead

    ×   Your previous content has been restored.   Clear editor

    ×   You cannot paste images directly. Upload or insert images from URL.


  • Similar Content

    • By Priyanka
      Science does not stand still and is constantly evolving. And one of its surprising innovations is the neural network. Thanks to them, people have simplified a large number of tasks that have taken on new technologies. And how it works, what they do, and whether we are threatened with the uprising of machines – learn from this material.
      What is a neural network

      A neural network is a system of interconnected processes working on a single task. It is able to work in a group and take into account previous experience and thus achieve great success
      Neural networks are an ordered system of processes interacting with each other. Each of them has to deal with incoming and outgoing signals. Signals are received and transmitted to each neuron, and this unites them into one network.
      Using the group mechanism of work, as well as a clearly defined algorithm that takes into account previous experience, neural networks can perform quite difficult and complex work. This makes them look like a human brain (albeit primitive)
      Painting neural networks

      Neural networks are able not only to draw themselves but also to combine 2 photos into one!
      Yes, now neural networks can act as full-fledged representatives in modern art. A striking example of this was shown by Stanford student Robert Barratt, who trained in neural network drawing.
      True, he started from another - first, by downloading about six thousand lines from Kanye West to a computer, he taught artificial intelligence how to write a rap. Then Robert downloaded a lot of different photos from the Internet. On their basis, the neural network and created a portrait of a fictional person.
      I must say that the picture turned out pretty low-quality, but for the first attempt it turned out well. Surprisingly, after a while, they were able to sell it at auction for a little less than half a million dollars.
      Car driving

      You may have seen similar cars: not only Google or Audi is testing the drones, but also, Yandex
      In the future, we hope that we will not have to drive ourselves. After all, this will create a system of unmanned vehicle control. And our super-technological neural networks will help us in this.
      At the moment, several large automotive companies like Audi or Tesla are engaged in similar developments. In the same area, they study the work of Uber and Google neural networks.
      With the help of a neural network system, a car without a driver can distinguish between road markings, signs, pedestrians and other cars, making a decision on how to pave the way.
      Scam Detection

      Very soon, the facts of corruption will be able to search and install not a person, but a computer!
      The main thing that initially sharpened neural networks – is the recognition of different images. In particular, they can display statistics and build a logical link between events. Or documents that were useful to Spanish scientists.
      On the basis of neural networks, they created a special program that through the analysis of various official documents can establish facts of corruption. This helps to control crime in many provincial regions.
      Also, neural networks help to recognize fraud with credit cards of people and are often used in the banking sector.
      Instant translation

      This is how Google translator works today - there is an instant auto-translation feature!
      Neural networks are now actively used by Google – one of the leading companies for the introduction of new technologies. And that's why they made a new function in their translator – instant translation of text in real-time using a special algorithm and neural networks.
      To do this, you just need to point the smartphone camera at the text. The neural network processes each character separately, even despite the poor quality, blur, sophisticated font. After the system quickly shifts the signs into words, connects the words into sentences, and the person instantly gets a ready translation.
      With such rapidly developing technologies, it is quite real people's fears that the uprising of machines may begin. But it seems that this is still far. The level of development of the most powerful network does not exceed the level of a five-year-old child. Yes, and we do not have large batteries, so it's enough to pull out the outlet to win 🙂
    • By DeZire
      Scientists from the University of California, San Diego managed to create samples of miniature brains that can form active neural networks. Despite its small size, artificial organs are able to emit similar to human brain waves at a certain stage of development.

      The pea-sized specimens, called "cerebral organoids," were created from human stem cells. After being placed in a special nutrient medium, they transformed into brain cells of several types and self-organized into complex 3D structures. About two months later, a team of researchers recorded bursts of brain activity of the organoid.
      The scientists used a machine-learning algorithm to compare the EEG characteristics of premature babies with the activity of cerebral organoids. According to Muotri, AI was able to determine the" age " of organoids, but between 25 and 40 weeks of development ceased to distinguish them from infants. For 10 months the researchers grew hundreds of samples of organelles. They also noted that miniature brains have "consciousness" - they are very rudimentary models.

      Biologists suggest that in the future, organoids can be used to study the development of human neural networks, modeling diseases, studying the evolution of the brain, monitoring drugs and even creating artificial intelligence.
  • Blogs

  • Categories

  • Gadgets


  • Reviews

    • OPPO Reno 3 review: Quad camera and lots of memory

      Last year's OPPO Reno 2 smartphone was remembered for its almost flagship characteristics. In the new generation, the emphasis has shifted - the role of the top-end device has been assigned to the Reno 3 Pro, and the standard “three” has become a mid-budget phone. However, Reno 3 received a powerful quad camera, high-quality screen, and a large amount of memory. We find out how good the new product is.

      Four rear modules and a 44-megapixel front-facing camera
      The main camera of Reno 3 received four modules, as in the top-end device of the series - Reno 3 Pro. However, there are some functions missing, such as interpolation up to 108 MP. All the blocks are arranged in a line on an islet that protrudes noticeably upwards if you wear the gadget without a cover.
      The main module has a 1/2-inch sensor with a resolution of 48 MP and a lens with an aperture of f/1.8. As usual, in standard mode, it produces 12-megapixel images, and if desired, you can take photos in full resolution - 48 MP. Telephoto is based on a 13-megapixel sensor and a lens with an f/2.4 aperture. The ultra-wide Angle is equipped with an 8-megapixel sensor and a lens with an f/2.2 aperture. Finally, the 2 MP sensor is designed to assess the depth of the scene.

      Telephoto, in addition to 2x optical zoom, provides a 5x hybrid zoom. The photo app allows you to quickly switch between four shooting modes: ultra-wide angle, 1x, 2x, and 5x.
      All modules can function together with HDR. This option works perfectly in Auto mode, improving both the light and dark tones of contrasting scenes. Artificial intelligence confidently recognizes the scenes being shot and selects settings.
      However, in non-standard lighting conditions, if you choose wide-angle optics, the color rendering will not always be accurate. This does not happen with the main module and the telephoto lens.

      The main module paired with the auxiliary module gives clear portraits and realistic blurs the background
      Reno 3 shoots well in mixed and artificial lighting, as well as in the absence of light in the dark. In the most problematic cases, the Night program helps out. It combines several shots and works out all the tones of contrasting scenes. For shooting with manual settings, an Expert preset is provided.
      Multi-megapixel selfies
      The front camera is built into the display and is located in the middle of its upper edge, in the "drop". The camera is equipped with a 44-megapixel sensor and a wide-angle f/2.4 aperture lens. The default settings are set to a 40-megapixel resolution of images. And the full one is selected by pressing a separate on-screen button. The image quality in both cases is approximately the same. Sharpness appears only from a distance of about 30 cm, that is, for example, it is impossible to obtain a macro photograph of your eye. It seems that the calculation was made for shooting with outstretched hands or a selfie stick. For self-portraits, there are traditionally many beauty options: smoothing the skin, correcting the shape of the face and nose, eye size, and so on.

      Selfies in 44MP mode
      The smartphone records 4K videos at 30 fps. The supported format is Full HD with 30 and 60 fps, as well as slow-motion clips in HD and Full HD. During movie shooting, you can freely switch the zoom.
      In 4K mode at 30 frames per second, the gadget produces a picture with the correct color reproduction and high definition. But when working with hands, stabilization is noticeably lacking.
      Shooting in Full HD at 60 fps uses Ultra Steady Video 2.0 stabilization, which makes the picture smoother.
      Work and entertainment without lags
      Reno 3 is built on an eight-core MediaTek Helio P90 chipset. The new product in AnTuTu gives out figures comparable to the results of Kirin 970 and Snapdragon 712, and slightly ahead of Snapdragon 710. And in the Geekbench 5 rating - is on par with Snapdragon 845. This speed is sufficient to avoid slowdowns in any applications.
      8 GB of RAM allows you to run multiple programs simultaneously. The built-in fingerprint sensor acts quickly and accurately. There are no performance problems in games. In the popular WoT: Blitz with default graphics settings, the frame rate is confidently kept at 60 fps.

      The device has only one main speaker, which produces clear sound without distortion up to about 90% of the volume. Plus, there is support for Dolby Atmos. In this software add-on, you can add surround sound and a bit of bass - a great option for multimedia applications.
      Unfortunately, you will not be able to create a full stereo pair, because the speaker is quieter than the main one. But in Reno3 Pro two speakers shake as if you are holding a good Bluetooth speaker in your hands. Complete in-ear headphones deliver high-quality sound. Thanks to the audio jack, you can connect your favorite accessory.

      The built-in storage is not too large: 128 GB. But it is easy to increase it by using a microSD memory card. It is placed in a triple slot next to two nano-SIM cards. We had a 512 GB test card at hand - it worked without problems.
      High autonomy
      The gadget is powered by a 4025 mAh battery and supports VOOC 3.0 fast charging technology. A 20 W adapter and a complete cable during the test provided recharging at high speed. For example, from 33 to 78%, the battery was charged in 13 minutes, and up to 100% in 27.
      An hour-long shootout in WoT: Blitz (display brightness - 50%) consumes 11 percent of the battery. That is, without going to the outlet, it is about 9 hours of offline gaming. In Flight-mode, with the same brightness, about 4.4% of the battery charge is spent per hour playing the video in Full HD. Therefore, you should expect 22 hours of offline video viewing. Excellent results!

      If long-term autonomy is not so important, you can disable power consumption optimization in the settings and select "High performance". In this case, the device will work a little faster, and the battery consumption will be stronger.
      Great screen
      Reno 3 is equipped with an AMOLED panel with a diagonal of 6.4 inches, a resolution of FHD+ (2400x1080 pixels), and an aspect ratio of 20:9. The display occupies 90.8% of the front panel area, has a teardrop-shaped cutout for the front and thin frames. The image is clear, with rich but natural colors. The viewing angles are large, and the brightness reserve is sufficient for using the phone on the street in any weather. With minimal screen light, it is comfortable to read in a dark room. And at night, you can switch to "Dark mode" - in fact, this is an inverse version of the design: with a black display and a white font. Many people will find it even more convenient, and it also saves battery life.
      Branded shell of the seventh version
      Reno 3 is running a proprietary ColorOS 7 shell based on Android 10. When comparing with the previous version, you can see that the font has not changed, but many icons have been redrawn. However, their shape and size can be easily changed at your own discretion. The screen can be converted to a split format - just make a three-finger swipe up.
      Specifications of OPPO Reno 3
      DISPLAY: 6.4 ", FHD + (2400x1080), 20: 9
      CHIPSET: CPU MediaTek Helio P90 (MT6779V), 2.2 GHz, GPU PowerVR GM 9446
      RAM: 8 GB
      STORAGE: 128 GB
      MAIN CAMERA: main - 48 megapixels (f/1.8) + 13 megapixels (telephoto) + 8 megapixels (wide-angle) + 2 megapixels (scene depth sensor);
      FRONT CAMERA: - 44 megapixels with f/2.4 aperture
      BATTERY: 4025 mAh
      FEATURES: a sub-screen fingerprint sensor, an NFC module, an audio output, support for microSD cards up to 256 GB (without SIM replacement), a proprietary shell ColorOS 7, three body colors: black, white and blue.
      Test results
      ANTUTU: 223,438
      GEEKBENCH: 5 428 single-core, 1568 multi-core
      AUTONOMY: about 22 hours of video playback

      Summary
      The OPPO Reno 3 is an interesting device with balanced characteristics. From the first minutes, the beautiful design, slim body, and high-quality 6.4-inch screen make you happy. Later on - high performance, fast recharging, and good battery life. The advantages include the photo-filling, including a 44-megapixel front camera. In addition, the built-in drive can be expanded with a memory card, but not to the detriment of the second SIM card - a trifle, but nice.

About Us

Worlds Tech News. CheapDeveloper provides the latest technology news, phone launch details, smartphone reviews, comparisons, laptops, cameras, Computer Hardware, Software, apps and more. You can trust our in-depth technology reviews to help you buy the best product for your needs.

×
×
  • Create New...