[step:Define the projection and prove that it is well-defined]
Define $\pi:E\to B$ by $\pi([(i,b,y)]):=b$. This definition is independent of the representative: if $(i,b,y)\sim(j,b',y')$, then the definition of $\sim$ gives $b=b'$. Therefore $\pi$ is well-defined.
For each $i\in I$, define $q_i:U_i\times F\to E$ by $q_i(b,y):=q(i,b,y)$.
Then
\begin{align*}
q_i(U_i\times F)=\pi^{-1}(U_i).
\end{align*}
Indeed, if $(b,y)\in U_i\times F$, then $\pi(q_i(b,y))=b\in U_i$, so $q_i(U_i\times F)\subseteq \pi^{-1}(U_i)$. Conversely, if $e\in \pi^{-1}(U_i)$ and $e=[(j,b,y')]$, then $b=\pi(e)\in U_i\cap U_j$. Since $\gamma_{ij}$ is a map from $(U_i\cap U_j)\times F$ to itself, there is a unique point $(b,y)\in (U_i\cap U_j)\times F$ such that
\begin{align*}
(b,y)=\gamma_{ij}(b,y').
\end{align*}
Thus $(i,b,y)\sim(j,b,y')$, so $e=q_i(b,y)$.
The map $q_i$ is injective. If $q_i(b,y)=q_i(b',y')$, then $(i,b,y)\sim(i,b',y')$. Hence $b=b'$ and
\begin{align*}
(b,y)=\gamma_{ii}(b,y')=(b,y'),
\end{align*}
so $y=y'$. Therefore $q_i$ is a bijection from $U_i\times F$ onto $\pi^{-1}(U_i)$.
[/step]