Women TICs in the entreprises
A Bolivian founder teaches a neural network to tell tigers from pandas
A folder of tiger and panda photos, a few lines of Python in Google Colab, and a neural network learns to label each picture. The same technique could sort clothes on an e-commerce marketplace by colour and price category. Abigail Geronimo, Founder-CEO of Bycicle and National Director General of ICT Women Bolivia, walks through the building blocks, then asks what it takes for women to turn such skills into startups.
Learn the TensorFlow and Keras concepts behind image classification, plus three startup tips
In short
Abigail Geronimo introduces TensorFlow and Keras for image classification, defines core concepts such as ReLU, softmax, flatten and Adam, and demonstrates a Google Colab notebook that labels pictures of tigers and pandas. She then links these skills to e-commerce and the growth of digital business in Latin America, closing with three startup recommendations: keep learning technology, build a strong team and never lose passion.
“startups very important you can change it every day you know the technology changes every day”
“the passion is very important don't forget uh technology is always implementing your startups”
What you will learn
- Explain what TensorFlow and Keras are and why they are used together for image classification.
- Define ReLU, softmax, flatten and the Adam optimizer in plain language.
- Identify the Python libraries (NumPy, OpenCV, Matplotlib, OS) used in a basic image classification notebook.
- Recognise how image classification can label e-commerce products by colour and price category.
- Apply three startup recommendations: keep learning technology, build a team with different skills and protect your passion.
Key takeaways from the speaker
- Please Add
Details
- Level
- Beginner
- Format
- Talk
- Technical depth
- Introductory technical
- For
- Student, Early career, Founder
- Length
- 19:30
- Published
- June 2021
Questions this talk answers
What is TensorFlow?
A free, open-source machine learning library created by Google, with a particular focus on training and inference of deep neural networks. It can be used together with Keras.
What do ReLU and softmax do in a neural network?
ReLU turns any value below zero into zero and returns positive values unchanged. Softmax converts a vector of numbers into a vector of probabilities proportional to each value's relative scale.
How can image classification be used in business?
E-commerce marketplaces can classify product pictures such as clothes, jeans and shoes by characteristics like colour and price category, so users see useful product information.
Why does the speaker see opportunity in Latin American digital business?
She points to fast e-commerce growth between 2018 and 2019 and a large rise in venture capital investment in the region, saying technology knowledge plus good marketing and finance can create new startups.
What does she recommend to women building startups?
Keep learning technology whatever you studied before, build a good team with different skills and professionals, keep innovating, and above all keep your passion.
About this talk
TensorFlow makes use of a graph framework. The graph gathers and describes all the series computations done during the training. In this talk about classified images with keras and Tensor flow to get label training.
Topics
Highlights
- ReLU, softmax and flatten in plain words (56 s)
- Why Latin America's digital business boom matters (73 s)
- Passion and a good team keep startups alive (76 s)
Speaker 1: Okay. It was nice to meet you, everyone. My name is Abigail Geronimo. I live in Bolivia. I'm from Bolivia. Nice to meet you, everyone, again. But about today, let's start with a magic classification with TensorFlow. Okay. Let's continue. But the first question is, what is TensorFlow? TensorFlow is created by Google. It's a free open source, a price library for mature learning. It's very, very useful, these tools. But if you now have developed these skills with TensorFlow and mature learners, it's cool. Congratulations.
Speaker 2: And that's, these, these tools, they can be used across a run-out text, but it has a particular focus and training on the inference of deep natural networks. For example, TensorFlow, it's a library based on data flow and different table programming. Speaker 2: And just with could be used with Keras. And you know, it's very, very important. I'm flexible tools, Keras with terms of flow.
Speaker 2: Okay. Okay. Today, in this very, a big global conference, 2021, in government tech, tell, let's start with this picture. Because you know, there are many, a picture of a tiger and panda. And there are also animals. And very interested. But it's very, very interesting with you. Speaker 2: It's very interesting with use terms of flow on and Keras, but trying it's different in narrow, narrow, narrow, and trying to different in machine learning, machine learning, you can see different pictures. If you know, there are many pixels, you can, you can classify with different.
um binary you know what's very important and that other other example this if you know there are the many e-commerce marketplace around the world on internet you know you can use this this skills winters of law and keras for example there are many pictures inside that is a data set there are many pictures about clothes jeans shows dress a t-shirts that you know where the internet you can sells on soul that's these products you know what is different a classified with price different price categories um colors pixel you know what's a characteristics colors and clarify what when this information to user final you know um so continue um i'll talk about about the open cv and by us matt palette list to their lister grid why because um after that i i'll show you uh a practice um
collab uh from google and um i prepare i practice and on python is very interesting okay but before a we we need to understand what's what's these words what's concept what's important about this practice for example um you know the computer vision is increasingly being used for imagine analysis and processing using artificial intelligent algorithms uh about the non-py for example um it's a library for uh for the python programming language that provides supports for creating large multimodational vectors and matrices along with a larger collection and high level mathematical function to operate of them and that's that's that's important on python and python uh for the python so um the python as a model allow with youtube performs you as the paints a operation such as creates a folders a list is a content for a
folders learning about the process terminating a process and so on it's very important what you implement um the model and python for example this one oh yes okay but in this practice we will uh we will smart blow it but what is model it okay uh what division is very interesting uh because um in this model it uh you know what the credit figure figure and one more and set effects returning and reference all the figure and the axis it's very important when you implement many many pictures and you can graphic you can show a very important the visual you know what gets a cool information for example uh which generates graph according to your wishes okay so um the next concept listed this one
what is a return list contains the names of the entries file of direction in the specific directory the reason we will use uh listed it okay but uh you know that l you you can you can listen too much all right but what is what means what is it's a structure that tired of that alone to restore an homogeneous set of data okay and all of them of the same type of the relate it's very important so another is for example um relu sort mounts for them what this is concept concept to implement in this practice uh and we will show you this it's a part of this practice um but relu is very important because you know uh when when you implement this type of practice and you need basically within a value x for
example if this value is less than zero if it converts it to serum i can feel it's a greater it returns the same value value this deep learning graph translate as whenever a neuron gives a negative output turn out that right neuro very important really the next concept softmax what is um softmax is mathematical function that converts a vector a number into a vector of probability where the probability of each value are professional to with their a relative scale of each value in the vector that reason we implement softmax in in this exercise okay and the use of the python efficient a flatten flatten what is flatten i don't know i don't understand okay no problem because a flattening use the scatter functions with drawing and match a and you need some data
for example you can first get the number of column and the required matrix in the then cover it to a single row matrix by calling and the flatten fusion for example and then call okay um why adam why what's the meaning i don't know okay no problem because other is um it's a extension uh of the gradients descent optimization algorithms if the reason we implement okay so uh continue i'll show you my exercise this one okay for example you know and you can see my screen uh i yes okay um so for example we need to implement um data sets okay very important implement data sets uh to the only and many many many many pictures you know but that's a for example for it could be uh close to the many what objects you know but that in this in this session in this
example we um use picture of all the tiger and panda okay uh okay i select the zip panda and tigers um so import c file uh import your data okay but you recognize in this exercise import cv2 numpy i told i told you what's meaning numpy before a map lead uh imagine you can you can it's very important what um determinates these concepts okay map lead and what's a determinates in this exercise with um in this collab of uh google okay okay in this in this session i i see the the what's a order art or position a currency in in this session about the matrix about the pictures you know it's very important and you know what's a position in that category in or select picture you know what the remaining picture of the same picture you just requires just one
of tiger what position uh what's a select a number you know but in this in this session i i'll see this okay okay um this one um you can see for uh for example grid because you know what's the category uh about the colors and you you can see the picture recognize the picture i select before okay and classify i i i see the number name with the number name with picture and what's the level with levels is very important where panda tigers there are the many pictures you know what's very beautiful because why why implement this okay i'll return my presentations now let's continue let's continue um you know um 100 billion in latin america for example was already experiencing exponential world and in commerce between
the 2018 and 2019 spending online shopping accepted uh the the startups enterprises there are many million millions around the world increasing incredible but the technology is for very important uh it's a factor very important around the world uh because you know the kobe 19 uh there are the many opportunity make different solution and make difference um get knowledge about the investigation when the improvements but continue uh a new model business a business a digital business you know what's very important but if you you know different skills um i don't know tool of all the microsoft um i don't know google google is very central because tensile flow um was celebrates um i don't know about the five years ago a birth continue modification continue uh growing you know what's about to implement this reality
um so between 2015 and 2019 venture capital investment in latin america have grown right a 673 percent and would stand at under core 11 and used for that six billions for the last years this information very important mission this conference because um there are other many money if you invest you can learn invest very good in good channel and in around the world about the business digital but if you get knowledge about implement this technology but you know use very good marketing and finances you can do many think of other startups new startups um growing startups to implement all in digital and i from i from bolivia i live in latin america i told you that a that information because it's very important implement latin america too much technology and it means very well about the business digital
i recommend this one because um it's very important when implementing startup with technology um but if you know if you um know different information about technology and trains around the world you can see more uh one more gets more uh opportunity in these challenge times okay startups very important you can change it every day you know the technology changes every day uh startups is very important implement in women and a good team you know when different skills different professionals different tools and but always innovate where do you um be okay technology uh you know very important uh always always always um is very important to show what's implement the value about the uh enterprise and in your uh enterprise and in your startups is the more most most important for me
is the passions passion is um it's an element very important for you because if you break in your startup um for many reason i don't know but if you continue get in its passions implements where whatever you um um live you can implement other startups when you get long edge before about the experience uh different startup i don't know but it's very important to is um get a good team uh very very very very important uh there are many startups implement uh in quarantine time and technology uh open the door are the many opportunities get you know and get knowledge get knowledge about different technology and all different language you know but if you prepare different um sessions but to continue a open mind in different uh aspect in you in your startup you can innovate more
um because uh the technology um every every country um have a close very very very very very very strong okay the passion is very important don't forget uh technology is always implementing your startups and startup always is very important when a woman leader our leader a shapes another another woman okay so it's very very important the three recommendation for you and not forget please the passion is the most important um okay okay uh i i if we have already uh i could i will come back in this exercise sorry in my exercise about this is if this exercise could be implement in marketplace e-commerce you know uh you know what the you there are many users it's important see that products these products price time weight
um colors um color so no i i i need see to buy it's very important okay so on others uh for me it's a picture to be here because i can share this information with you uh but i want to motivate you uh because it's very very important there's very important there are many enterprises um leaders by women okay so sorry okay okay okay not forget okay um i repeat you a birth is very important it creates uh a startup in this challenge time um technology not forget um get knowledge about a technology um it doesn't matter what corporate you've studied before okay technology always always is very important to get knowledge in you and i don't know it's many personal uh say um the technology is my hobby i don't know i i heard out
something like that okay it is the most important is the passions not forget please okay uh okay um my name is abigail jeronimo i nice to meet you nice to meet you i'm sorry my apply about my english sorry but my first language is espanol spanish and you can find in this github twitter linkedin um abigail jeronimo thank you everyone thank you the team uh woman take congratulations about this global conference you can find in this global conference 2021 thank you so much you
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Frameworks and ideas mentioned
- ReLU activation
- Softmax
- Adam optimizer
- Gradient descent
Resources mentioned
- TensorFlow
- Keras
- Google Colab
- NumPy
- OpenCV
- Matplotlib
Cite this talk
Abigail Geronimo. "Women TICs in the entreprises." WomenTech Network, 2021. https://www.womentech.net/video/abigail-geronimo-women-tics-in-entreprises