Some brilliant mind who lurks on HN should create a competitor to eBay. Not some online yard sale where you have to meet people in person for the exchange, but a nice online marketplace for people to sell their stuff.<p>If this already exists, do tell.<p>In my opinion, eBay for sellers (outside of maybe the power sellers. I've sold <100 items) is a nightmare.<p>I sold a Tiffany necklace a few months back. The buyer reported to eBay it was fake and I was ordered to refund the seller and they could keep the fake item. Long story short, over a span of a few weeks, I was lucky enough to get an original receipt from Tiffany and supply it as evidence, which almost still did not work.<p>More recently, I listed an old iPhone with a "buy it now". Sold in like 30 minutes to someone with an "@god[dot].com" email requesting I send the phone first and then he will pay me. It took a week or longer for me to challenge this. I even have settings to disallow certain types of eBay users based on ratings etc. eBay still charged me a listing fee.<p>I had success doing a non buy it now sale after that. Maybe that is my only option now.<p>I don't even feel like getting into how slow their seller admin tools are ...
I thought the title said "let you STOP using photos (when selling)", which is what has basically already happened due to the prevalence of stock images and CGI items on a white background.<p>I'm still looking for the browser extension that lets you filter out any listing with a pure white or gradient background from your eBay, etsy, pinterest, etc. searches.<p>Would be a HUGE indicator of an actual real used item sold by a human, vs. a new dropshipped item from China. You know, what I want to use eBay to buy.
I'm pretty sure Taobao in China has had this for ages. Search suggests[0] since as early as 2011. To test it, go to Taobao[1] and click the photo icon at the right of the search box (previously only available in some countries and product categories, now apparently global). If this was the other way around it would be "OMG China is copying the west!" No such discussion the other way around. Just sayin'...<p>[0] <a href="https://www.chinainternetwatch.com/1189/taobao-imagine-an-image-search-engine-for-shoppers/" rel="nofollow">https://www.chinainternetwatch.com/1189/taobao-imagine-an-im...</a><p>[1] <a href="http://www.taobao.com/" rel="nofollow">http://www.taobao.com/</a>
I have been trying to use Amazon's visual search for the past week or two and out of many trials, it has only worked once.<p>I wonder whether eBay will beat Amazon in this category as the types of photos they have access to (amateur/home vs. professional/stock) are very different. Either way, I am not confident that taking a picture is a better solution than just typing in the name for the vast majority of items.
My pet startup idea, going back I suppose 15+ years is a service that does this:<p>You have something. You like it. You want another one. The service gets it for you. This doesn't have to work from a picture -- it could use a UPC barcode or a text description.<p>Idea originally inspired by a school friends of mine who wore the exact same type of boots for years and years. When the old pair wore out, he'd buy another pair the exact same.<p>Over the years I've noticed many instances where I really want another one of something I have but it turns out to be either painful or impossible to find. You know however that somewhere someone has a warehouse full of the things.<p>Also like the electronics industry brokers who can track down an old component for you to repair gear or re-start manufacturing on some old product. Those have existed since at least the 1970's.<p>A few years ago I tested Amazon's visual product search feature to see if it would do what I wanted. It didn't. I showed an image of some brown boots I was wearing, and certainly the suggested products were also brown boots, but not the same kind.
It seems like ebay would have a ton of user-provided photos which would be a possibly <i>huge</i> corpus for machine learning to help identify objects like this. It will be interesting to see how well this works!
I know a bunch of sites has this feature already but I'm pretty excited to try it out since eBay is probably the website with the best dataset. I do not think any other site has as many home taken pictures as eBay. This would definitely make the image recognition much more robust. I'm guessing the problem with Amazon or TaoBao is that most of the pictures are studio pictures and does not scale well when a person takes a picture in a random environment.<p>That being said though, the eBay buying and selling experience is a hit or miss. It sucks how there is group of people preying on the site to rip people off. I had an selling experience so bad that I vouched never to sell anything via eBay anymore. Craiglist tends to be a much better transaction.
To test a machine learning algorithm, one can use "Fashion-MNIST": <a href="https://github.com/zalandoresearch/fashion-mnist" rel="nofollow">https://github.com/zalandoresearch/fashion-mnist</a><p>"Fashion-MNIST is a dataset of Zalando's article images—consisting of a training set of 60,000 examples and a test set of 10,000 examples. Each example is a 28x28 grayscale image, associated with a label from 10 classes. We intend Fashion-MNIST to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms. It shares the same image size and structure of training and testing splits."
Seems like a great way for eBay to undercut brick and mortar stores (and for consumers to save money) on cheap junk from China that people would rather pay $0.99 and wait 3wk for than pay $4.99 and have right now.
I have made a browser extension[1][2] for those looking to perform reverse image searches with multiple engines at once. It's more reliable than Chrome's built-in image search, which appears in the context menu only for <img> elements, while this extension parses the clicked area to extract images. Support for shopping sites may come in the future.<p>[1] <a href="https://chrome.google.com/webstore/detail/search-by-image/cnojnbdhbhnkbcieeekonklommdnndci" rel="nofollow">https://chrome.google.com/webstore/detail/search-by-image/cn...</a><p>[2] <a href="https://addons.mozilla.org/en-US/firefox/addon/search_by_image/" rel="nofollow">https://addons.mozilla.org/en-US/firefox/addon/search_by_ima...</a>
Interesting how we built a raw prototype about eight months ago (<a href="https://imgur.com/OIaQu7x" rel="nofollow">https://imgur.com/OIaQu7x</a>), but it never took off (at least for the customer whom we've pitched).
To add to other's comments, the visual search can be quite useful comparing to terms search specifically for non-native speakers and people who are not skilled in fashion terms (i.e., the name of the material, pattern).
As for how accurate it is, this is indeed an existing problem. If you product set is significant (>10k products), then the visual search results can be entirely off the target, not surprised Ebay is dealing with the same problem.
Can someone explain to me why eBay is trading only at 5.5x P/E? I fully understand that it isn't growing tremendously but with all its classified assets, I can't help but feel that it should be trading at a much higher market cap.
Last weekend, my dad needed new printer cartridges and so we went sifting through the internet for different offerings. By chance, we got onto a page on the manufacturer's website, which had images of all their cartridge packs. Among those, we also found an XL version of what we initially thought to buy.<p>So, there we were, needing exactly this feature. If I didn't know that it's impossible for them to have developed this in such a short time, I'd be a little freaked.<p>Thankfully, though, TinEye served the exact same purpose...
Can anyone recommend a good tutorial or reading on how to setup content-based (visual) image searching using a CNN to process images. I'm looking to build a POC of a reverse image search trained with in-house product data. In the past I've used the imgSeek but it is dated and not using neural nets.
We provide similar technology as a service if anyone is interested to hire us. Check our demo at: <a href="https://app.turingiq.com/demo/image/client" rel="nofollow">https://app.turingiq.com/demo/image/client</a>
Shameless plug for my friend's startup, which has prototyped the concept and got a few customers for the last year.<p><a href="https://www.mirrorthatlook.com/" rel="nofollow">https://www.mirrorthatlook.com/</a>
It would be interesting if they'd use this tech to deduplicate the gazillion listings for many cheap (chinese) goods which seem to use common set of images with maybe varying watermarks etc.